Modeling_and_Interpreting.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "accelerator": "GPU",
    "colab": {
      "name": "Modeling_and_Interpreting.ipynb",
      "provenance": [],
      "collapsed_sections": [],
      "toc_visible": true,
      "machine_shape": "hm",
      "mount_file_id": "1fasbo4TM3VyhL8ZkUYap0v6rsVNJXWxA",
      "authorship_tag": "ABX9TyNMw3/GTkhz8IHshddBLb4d",
      "include_colab_link": true
    },
    "kernelspec": {
      "display_name": "Python 3",
      "name": "python3"
    }
  },
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "view-in-github",
        "colab_type": "text"
      },
      "source": [
        "<a href=\"https://colab.research.google.com/github/AvonleaFisher/Analyzing-NYC-311-Service-Requests/blob/main/Modeling_and_Interpreting.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "BuPIppTCJQoQ"
      },
      "source": [
        "# Introduction "
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "zalxhpvEJQld"
      },
      "source": [
        "The aim of this section is to train a neural network that can accurately classify the agency that responded to a 311 call. In the exploratory analysis section, we saw that none of the community districts' numeric indicators were highly correlated with any of the dummy-encoded agency categories. Intuitively, one may expect the call's descriptor to be the best predictor of the responding agency, but given that the descriptors consist of natural language data, they will need to undergo additional preprocessig before being fed into a model."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "GoZ-JIQ3cb2u"
      },
      "source": [
        "from tensorflow.random import set_seed\n",
        "set_seed(9)\n",
        "import numpy as np\n",
        "np.random.seed(999)"
      ],
      "execution_count": 191,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "EQ2JuRRPUfI6"
      },
      "source": [
        "# Loading Dependencies"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "muoTtQYo0uhz"
      },
      "source": [
        "!pip install -U -q PyDrive\n",
        "from pydrive.auth import GoogleAuth\n",
        "from pydrive.drive import GoogleDrive\n",
        "from google.colab import auth\n",
        "from oauth2client.client import GoogleCredentials\n",
        "from sklearn.model_selection import train_test_split\n",
        "from sklearn.metrics import confusion_matrix\n",
        "\n",
        "\n",
        "import matplotlib.pyplot as plt\n",
        "import pandas as pd\n",
        "import numpy as np\n",
        "import seaborn as sns\n",
        "import random\n",
        "import re\n",
        "\n",
        "from sklearn import preprocessing\n",
        "import tensorflow as tf\n",
        "from tensorflow import keras\n",
        "from tensorflow.keras import layers\n",
        "\n",
        "from keras.preprocessing.text import Tokenizer\n",
        "from keras.preprocessing.sequence import pad_sequences\n",
        "from keras import models\n",
        "from keras import layers\n",
        "from keras import optimizers\n",
        "from keras.utils.np_utils import to_categorical\n",
        "from keras.models import Sequential\n",
        "from tensorflow.keras.layers import Embedding, LSTM, Dense, Activation, Dropout"
      ],
      "execution_count": 192,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "8taQeJ4zZwZu"
      },
      "source": [
        "# Importing the Data\n",
        "The data used in this notebook can be accessed via the following sharable Google Drive link:\n",
        "\n",
        "[df_final.csv](https://drive.google.com/file/d/1Jqxrz8lDY_iYYur7eJaBt9fmf-O2UqLO/view?usp=sharing)\n"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "dHtsa82d0yeh"
      },
      "source": [
        "auth.authenticate_user()\n",
        "gauth = GoogleAuth()\n",
        "gauth.credentials = GoogleCredentials.get_application_default()\n",
        "drive = GoogleDrive(gauth)"
      ],
      "execution_count": 193,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "WZHghdba0oB5",
        "outputId": "31ce5044-35a7-46c1-d7c0-6ed38b733e6a"
      },
      "source": [
        "path = '/content/drive/MyDrive/Colab Notebooks/df_final.csv' #change path to match your Google Drive file location\n",
        "df = pd.read_csv(path, index_col=0)"
      ],
      "execution_count": 194,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "/usr/local/lib/python3.6/dist-packages/numpy/lib/arraysetops.py:569: FutureWarning: elementwise comparison failed; returning scalar instead, but in the future will perform elementwise comparison\n",
            "  mask |= (ar1 == a)\n"
          ],
          "name": "stderr"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "q395NFX9yEMQ"
      },
      "source": [
        "# Assessing Model Performance"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "9LUNQFdHyb4p"
      },
      "source": [
        "With perfectly balanced classes, a model's accuracy is acceptable if it exceeds the expected accuracy rate of simply picking a random class. With heavily imbalanced classes, however, a model's accuracy can instead be compared to that of picking the majority class. Below is the expected accuracy rate if we were to predict that all calls in a sample were responded to by the NYPD."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "DNrao_hK2sFG",
        "outputId": "abd8e004-a5ac-42ad-d35e-64e70cd91e83"
      },
      "source": [
        "#divide count of NYPD calls by total calls and round to 2 decimal places\n",
        "round(len(df[df.agency =='NYPD'])/len(df) * 100, 2) "
      ],
      "execution_count": 195,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "52.61"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 195
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "1TpY--8t3X-b"
      },
      "source": [
        "While the aim will be to maximize model accuracy, any classifier that is at least 53% accurate will out-perform the simple assignment of all calls to the NYPD."
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "W9fRwJwwMgbA"
      },
      "source": [
        "# Sampling the Data \n",
        "The dataset currently consists of more than one million rows:"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "mMMJh58Y8dIA",
        "outputId": "ce9f9853-0748-411b-ea1c-b1a5128609fa"
      },
      "source": [
        "len(df)"
      ],
      "execution_count": 196,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "1379786"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 196
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "JMUcv--x8aaY"
      },
      "source": [
        "Retrieving the training data from a representative sample, rather than the entire dataset, will save time in the training process and ideally not result in a significant decline in accuracy.\n",
        "\n",
        "In training classification models, we'll want to ensure that complaint descriptors from every agency category are represented. This way, the model will also be able to recognize encoded descriptors that correspond to agencies that rarely receive calls. However, training a model on samples of equal length from each agency may lead to confusion when it is presented with new, imbalanced data. We'll oversample the minority classes, but ensure that the largest samples are those agencies that receive the most calls. This will allow the diversity of descriptors associated with each agency to be captured in the samples. Below, we create and print a dictionary with the class weights."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "m2JjrNL748aM",
        "outputId": "8838c1c2-a3ec-4f39-89a0-cb4ad0088ce8"
      },
      "source": [
        "#create dictionary to store weights\n",
        "weights_dict = {}\n",
        "\n",
        "#loop through agency: value count pairs\n",
        "for agency, count in list(zip(df.agency.value_counts().index, \n",
        "                              df.agency.value_counts().values)):\n",
        "#calculate weights and update dictionary \n",
        "  weight = count/len(df)\n",
        "  weights_dict[agency] = weight\n",
        "weights_dict"
      ],
      "execution_count": 197,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "{'DCA': 0.007849768007502613,\n",
              " 'DEP': 0.057544430803037575,\n",
              " 'DOB': 0.031322248522596983,\n",
              " 'DOE': 0.0003138167802833193,\n",
              " 'DOHMH': 0.022252001397318134,\n",
              " 'DOITT': 0.00011233626084044916,\n",
              " 'DOT': 0.059362103978443034,\n",
              " 'DPR': 0.08357020581452486,\n",
              " 'DSNY': 0.04376330822315924,\n",
              " 'EDC': 0.004157891151236496,\n",
              " 'HPD': 0.12691315899711986,\n",
              " 'NYPD': 0.5261301390215584,\n",
              " 'OSE': 0.03320152545394721,\n",
              " 'TLC': 0.0035070655884318295}"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 197
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "uPPBjrs5LGG6",
        "outputId": "19e4adf9-5c58-47a1-c3ed-273880bba3ac"
      },
      "source": [
        "df.agency.value_counts()"
      ],
      "execution_count": 198,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "NYPD     725947\n",
              "HPD      175113\n",
              "DPR      115309\n",
              "DOT       81907\n",
              "DEP       79399\n",
              "DSNY      60384\n",
              "OSE       45811\n",
              "DOB       43218\n",
              "DOHMH     30703\n",
              "DCA       10831\n",
              "EDC        5737\n",
              "TLC        4839\n",
              "DOE         433\n",
              "DOITT       155\n",
              "Name: agency, dtype: int64"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 198
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "gh_WXYsAPfWm"
      },
      "source": [
        "Next, we'll create a new dictionary with guidelines for the number of samples to be drawn from each class. The class weights are multiplied by a term that results in the largest sample, NYPD, consisting of around 60k calls."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "Eei-2kEy9iCS"
      },
      "source": [
        "n_samples_dict = {}\n",
        "\n",
        "#loop through agency: weight pairs\n",
        "for agency, weight in zip(weights_dict.keys(), weights_dict.values()):\n",
        "\n",
        "#multiply each weight by 15**4.3 and update the dictionary\n",
        "  n_samples = int(15**4.3 * weight)\n",
        "  n_samples_dict[agency] = n_samples"
      ],
      "execution_count": 199,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "qE__5eWsBH0c",
        "outputId": "908df560-06ea-4bba-b217-518f1e7cdefa"
      },
      "source": [
        "n_samples_dict"
      ],
      "execution_count": 200,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "{'DCA': 895,\n",
              " 'DEP': 6564,\n",
              " 'DOB': 3573,\n",
              " 'DOE': 35,\n",
              " 'DOHMH': 2538,\n",
              " 'DOITT': 12,\n",
              " 'DOT': 6771,\n",
              " 'DPR': 9533,\n",
              " 'DSNY': 4992,\n",
              " 'EDC': 474,\n",
              " 'HPD': 14477,\n",
              " 'NYPD': 60018,\n",
              " 'OSE': 3787,\n",
              " 'TLC': 400}"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 200
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "1g8WnjUsRmw2"
      },
      "source": [
        "The below function returns weighted samples of an agency that's passed in."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "geHzoyWv52OZ"
      },
      "source": [
        "def get_weighted_sample(agency):\n",
        "  \n",
        "  \"\"\"Takes in an agency name as a string, and creates a subset with all rows\n",
        "     with a match in the agency column. Returns a weighted sample of call data\n",
        "     associated with that agency, based on the values in n_samples_dict.\"\"\"\n",
        "\n",
        "  agency_subset = df[df.agency == agency]\n",
        "\n",
        "  n_samples = n_samples_dict[agency]\n",
        "  \n",
        "  weighted_sample = agency_subset.sample(n=n_samples, random_state=9)\n",
        "\n",
        "  return weighted_sample"
      ],
      "execution_count": 201,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "__2PdUBESxkK"
      },
      "source": [
        "The function is applied below to each unique agency to create a dataframe of samples, and the dataframe is appended to the list 'samples.' These are conatenated in 'weighted_df' and the index is reset."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "uR6-bbBC-MiY"
      },
      "source": [
        "samples = []\n",
        "\n",
        "for agency in df.agency.unique():\n",
        "  samples.append(get_weighted_sample(agency))"
      ],
      "execution_count": 202,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "ny-Ss-YVFcCJ"
      },
      "source": [
        "weighted_df = pd.concat(samples)\n",
        "weighted_df.reset_index(inplace=True)"
      ],
      "execution_count": 203,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 760
        },
        "id": "-OP7tnoiFbSk",
        "outputId": "d17bbff6-cc61-4773-eccd-08fb5e5d5489"
      },
      "source": [
        "weighted_df"
      ],
      "execution_count": 204,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/html": [
              "<div>\n",
              "<style scoped>\n",
              "    .dataframe tbody tr th:only-of-type {\n",
              "        vertical-align: middle;\n",
              "    }\n",
              "\n",
              "    .dataframe tbody tr th {\n",
              "        vertical-align: top;\n",
              "    }\n",
              "\n",
              "    .dataframe thead th {\n",
              "        text-align: right;\n",
              "    }\n",
              "</style>\n",
              "<table border=\"1\" class=\"dataframe\">\n",
              "  <thead>\n",
              "    <tr style=\"text-align: right;\">\n",
              "      <th></th>\n",
              "      <th>index</th>\n",
              "      <th>created_date</th>\n",
              "      <th>complaint_type</th>\n",
              "      <th>agency</th>\n",
              "      <th>community_board</th>\n",
              "      <th>latitude</th>\n",
              "      <th>longitude</th>\n",
              "      <th>descriptor</th>\n",
              "      <th>borough</th>\n",
              "      <th>day</th>\n",
              "      <th>month</th>\n",
              "      <th>day_of_week</th>\n",
              "      <th>hour</th>\n",
              "      <th>count_hosp_clinic</th>\n",
              "      <th>count_libraries</th>\n",
              "      <th>count_parks</th>\n",
              "      <th>count_public_schools</th>\n",
              "      <th>crime_per_1000</th>\n",
              "      <th>female_20_24</th>\n",
              "      <th>female_25_29</th>\n",
              "      <th>female_30_34</th>\n",
              "      <th>female_35_39</th>\n",
              "      <th>female_40_44</th>\n",
              "      <th>female_45_49</th>\n",
              "      <th>female_50_54</th>\n",
              "      <th>female_55_59</th>\n",
              "      <th>female_60_64</th>\n",
              "      <th>male_15_19</th>\n",
              "      <th>male_20_24</th>\n",
              "      <th>male_30_34</th>\n",
              "      <th>male_40_44</th>\n",
              "      <th>male_45_49</th>\n",
              "      <th>male_50_54</th>\n",
              "      <th>male_55_59</th>\n",
              "      <th>male_60_64</th>\n",
              "      <th>under18_rate</th>\n",
              "      <th>over65_rate</th>\n",
              "      <th>lots_total</th>\n",
              "      <th>lots_vacant</th>\n",
              "      <th>pct_white_nh</th>\n",
              "      <th>pct_black_nh</th>\n",
              "      <th>pct_asian_nh</th>\n",
              "      <th>pct_other_nh</th>\n",
              "      <th>pct_hispanic</th>\n",
              "      <th>mean_commute</th>\n",
              "      <th>pop_change_00_10</th>\n",
              "      <th>pct_clean_strts</th>\n",
              "      <th>unemployment</th>\n",
              "      <th>poverty_rate</th>\n",
              "      <th>lep_rate_boro</th>\n",
              "      <th>pct_hh_rent_burd</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <th>0</th>\n",
              "      <td>383700</td>\n",
              "      <td>2020-10-02 10:07:39</td>\n",
              "      <td>General Construction/Plumbing</td>\n",
              "      <td>DOB</td>\n",
              "      <td>7 BROOKLYN</td>\n",
              "      <td>40.638071</td>\n",
              "      <td>-74.007711</td>\n",
              "      <td>working contrary to stop work order</td>\n",
              "      <td>BROOKLYN</td>\n",
              "      <td>2020-10-02</td>\n",
              "      <td>10</td>\n",
              "      <td>4</td>\n",
              "      <td>10</td>\n",
              "      <td>32</td>\n",
              "      <td>2</td>\n",
              "      <td>4</td>\n",
              "      <td>21</td>\n",
              "      <td>8.3</td>\n",
              "      <td>3.0</td>\n",
              "      <td>4.9</td>\n",
              "      <td>4.6</td>\n",
              "      <td>4.2</td>\n",
              "      <td>3.4</td>\n",
              "      <td>3.4</td>\n",
              "      <td>3.2</td>\n",
              "      <td>2.4</td>\n",
              "      <td>2.4</td>\n",
              "      <td>2.3</td>\n",
              "      <td>2.8</td>\n",
              "      <td>5.0</td>\n",
              "      <td>3.8</td>\n",
              "      <td>3.3</td>\n",
              "      <td>3.3</td>\n",
              "      <td>2.6</td>\n",
              "      <td>2.2</td>\n",
              "      <td>24.0</td>\n",
              "      <td>10.3</td>\n",
              "      <td>13315</td>\n",
              "      <td>177</td>\n",
              "      <td>24.4</td>\n",
              "      <td>2.6</td>\n",
              "      <td>31.3</td>\n",
              "      <td>1.9</td>\n",
              "      <td>39.8</td>\n",
              "      <td>42.9</td>\n",
              "      <td>0.05</td>\n",
              "      <td>92.6</td>\n",
              "      <td>3.8</td>\n",
              "      <td>27.9</td>\n",
              "      <td>23.6</td>\n",
              "      <td>48.1</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>1</th>\n",
              "      <td>164754</td>\n",
              "      <td>2020-09-15 11:47:20</td>\n",
              "      <td>Elevator</td>\n",
              "      <td>DOB</td>\n",
              "      <td>4 MANHATTAN</td>\n",
              "      <td>40.752111</td>\n",
              "      <td>-74.004645</td>\n",
              "      <td>elevator  multiple devices on property</td>\n",
              "      <td>MANHATTAN</td>\n",
              "      <td>2020-09-15</td>\n",
              "      <td>9</td>\n",
              "      <td>1</td>\n",
              "      <td>11</td>\n",
              "      <td>38</td>\n",
              "      <td>2</td>\n",
              "      <td>7</td>\n",
              "      <td>29</td>\n",
              "      <td>18.1</td>\n",
              "      <td>4.0</td>\n",
              "      <td>7.2</td>\n",
              "      <td>6.4</td>\n",
              "      <td>3.8</td>\n",
              "      <td>3.0</td>\n",
              "      <td>2.9</td>\n",
              "      <td>2.3</td>\n",
              "      <td>2.7</td>\n",
              "      <td>2.7</td>\n",
              "      <td>1.1</td>\n",
              "      <td>3.3</td>\n",
              "      <td>7.2</td>\n",
              "      <td>4.8</td>\n",
              "      <td>3.9</td>\n",
              "      <td>3.6</td>\n",
              "      <td>3.7</td>\n",
              "      <td>3.1</td>\n",
              "      <td>8.2</td>\n",
              "      <td>13.4</td>\n",
              "      <td>3417</td>\n",
              "      <td>70</td>\n",
              "      <td>59.1</td>\n",
              "      <td>5.2</td>\n",
              "      <td>17.2</td>\n",
              "      <td>3.0</td>\n",
              "      <td>15.5</td>\n",
              "      <td>26.7</td>\n",
              "      <td>0.18</td>\n",
              "      <td>95.6</td>\n",
              "      <td>3.5</td>\n",
              "      <td>11.3</td>\n",
              "      <td>15.8</td>\n",
              "      <td>33.9</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>2</th>\n",
              "      <td>1301458</td>\n",
              "      <td>2020-07-02 15:02:08</td>\n",
              "      <td>General Construction/Plumbing</td>\n",
              "      <td>DOB</td>\n",
              "      <td>6 BROOKLYN</td>\n",
              "      <td>40.676215</td>\n",
              "      <td>-74.008818</td>\n",
              "      <td>building shaking vibrating structural stability</td>\n",
              "      <td>BROOKLYN</td>\n",
              "      <td>2020-07-02</td>\n",
              "      <td>7</td>\n",
              "      <td>3</td>\n",
              "      <td>15</td>\n",
              "      <td>18</td>\n",
              "      <td>4</td>\n",
              "      <td>8</td>\n",
              "      <td>24</td>\n",
              "      <td>7.8</td>\n",
              "      <td>2.4</td>\n",
              "      <td>5.3</td>\n",
              "      <td>6.7</td>\n",
              "      <td>5.6</td>\n",
              "      <td>4.3</td>\n",
              "      <td>3.5</td>\n",
              "      <td>2.8</td>\n",
              "      <td>2.6</td>\n",
              "      <td>2.3</td>\n",
              "      <td>1.8</td>\n",
              "      <td>1.3</td>\n",
              "      <td>5.8</td>\n",
              "      <td>4.4</td>\n",
              "      <td>3.9</td>\n",
              "      <td>2.7</td>\n",
              "      <td>2.4</td>\n",
              "      <td>2.0</td>\n",
              "      <td>21.2</td>\n",
              "      <td>10.4</td>\n",
              "      <td>13799</td>\n",
              "      <td>406</td>\n",
              "      <td>64.1</td>\n",
              "      <td>7.2</td>\n",
              "      <td>7.0</td>\n",
              "      <td>5.4</td>\n",
              "      <td>16.3</td>\n",
              "      <td>38.4</td>\n",
              "      <td>0.01</td>\n",
              "      <td>95.9</td>\n",
              "      <td>4.0</td>\n",
              "      <td>9.6</td>\n",
              "      <td>23.6</td>\n",
              "      <td>28.8</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>3</th>\n",
              "      <td>661127</td>\n",
              "      <td>2020-08-29 15:41:01</td>\n",
              "      <td>Building/Use</td>\n",
              "      <td>DOB</td>\n",
              "      <td>12 QUEENS</td>\n",
              "      <td>40.696722</td>\n",
              "      <td>-73.761430</td>\n",
              "      <td>illegal conversion of residential building space</td>\n",
              "      <td>QUEENS</td>\n",
              "      <td>2020-08-29</td>\n",
              "      <td>8</td>\n",
              "      <td>5</td>\n",
              "      <td>15</td>\n",
              "      <td>34</td>\n",
              "      <td>9</td>\n",
              "      <td>12</td>\n",
              "      <td>46</td>\n",
              "      <td>11.2</td>\n",
              "      <td>3.6</td>\n",
              "      <td>4.1</td>\n",
              "      <td>3.6</td>\n",
              "      <td>3.6</td>\n",
              "      <td>3.2</td>\n",
              "      <td>3.8</td>\n",
              "      <td>4.0</td>\n",
              "      <td>3.7</td>\n",
              "      <td>3.2</td>\n",
              "      <td>3.0</td>\n",
              "      <td>3.5</td>\n",
              "      <td>3.4</td>\n",
              "      <td>2.8</td>\n",
              "      <td>3.0</td>\n",
              "      <td>3.5</td>\n",
              "      <td>3.1</td>\n",
              "      <td>2.7</td>\n",
              "      <td>22.4</td>\n",
              "      <td>13.3</td>\n",
              "      <td>41418</td>\n",
              "      <td>1189</td>\n",
              "      <td>1.8</td>\n",
              "      <td>61.8</td>\n",
              "      <td>12.6</td>\n",
              "      <td>7.9</td>\n",
              "      <td>15.9</td>\n",
              "      <td>49.0</td>\n",
              "      <td>0.01</td>\n",
              "      <td>94.9</td>\n",
              "      <td>5.9</td>\n",
              "      <td>19.0</td>\n",
              "      <td>28.8</td>\n",
              "      <td>46.9</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>4</th>\n",
              "      <td>252987</td>\n",
              "      <td>2020-06-04 13:20:26</td>\n",
              "      <td>Covid-19 Non-Essential Construction</td>\n",
              "      <td>DOB</td>\n",
              "      <td>11 BROOKLYN</td>\n",
              "      <td>40.600670</td>\n",
              "      <td>-73.982837</td>\n",
              "      <td>construction</td>\n",
              "      <td>BROOKLYN</td>\n",
              "      <td>2020-06-04</td>\n",
              "      <td>6</td>\n",
              "      <td>3</td>\n",
              "      <td>13</td>\n",
              "      <td>9</td>\n",
              "      <td>2</td>\n",
              "      <td>2</td>\n",
              "      <td>21</td>\n",
              "      <td>5.4</td>\n",
              "      <td>2.8</td>\n",
              "      <td>3.8</td>\n",
              "      <td>3.9</td>\n",
              "      <td>3.5</td>\n",
              "      <td>3.6</td>\n",
              "      <td>3.4</td>\n",
              "      <td>3.5</td>\n",
              "      <td>3.3</td>\n",
              "      <td>3.1</td>\n",
              "      <td>2.4</td>\n",
              "      <td>2.9</td>\n",
              "      <td>3.9</td>\n",
              "      <td>3.4</td>\n",
              "      <td>3.2</td>\n",
              "      <td>3.4</td>\n",
              "      <td>3.3</td>\n",
              "      <td>2.8</td>\n",
              "      <td>21.3</td>\n",
              "      <td>15.6</td>\n",
              "      <td>21821</td>\n",
              "      <td>150</td>\n",
              "      <td>40.2</td>\n",
              "      <td>1.0</td>\n",
              "      <td>40.8</td>\n",
              "      <td>2.1</td>\n",
              "      <td>15.9</td>\n",
              "      <td>46.6</td>\n",
              "      <td>0.06</td>\n",
              "      <td>93.9</td>\n",
              "      <td>4.1</td>\n",
              "      <td>22.5</td>\n",
              "      <td>23.6</td>\n",
              "      <td>49.9</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>...</th>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>114064</th>\n",
              "      <td>1003068</td>\n",
              "      <td>2020-09-29 14:16:21</td>\n",
              "      <td>Linknyc</td>\n",
              "      <td>DOITT</td>\n",
              "      <td>1 MANHATTAN</td>\n",
              "      <td>40.715836</td>\n",
              "      <td>-74.002832</td>\n",
              "      <td>damaged defective</td>\n",
              "      <td>MANHATTAN</td>\n",
              "      <td>2020-09-29</td>\n",
              "      <td>9</td>\n",
              "      <td>1</td>\n",
              "      <td>14</td>\n",
              "      <td>10</td>\n",
              "      <td>2</td>\n",
              "      <td>13</td>\n",
              "      <td>19</td>\n",
              "      <td>6.8</td>\n",
              "      <td>3.9</td>\n",
              "      <td>7.2</td>\n",
              "      <td>6.5</td>\n",
              "      <td>4.7</td>\n",
              "      <td>3.6</td>\n",
              "      <td>2.9</td>\n",
              "      <td>2.3</td>\n",
              "      <td>2.4</td>\n",
              "      <td>2.6</td>\n",
              "      <td>1.8</td>\n",
              "      <td>2.8</td>\n",
              "      <td>6.3</td>\n",
              "      <td>3.6</td>\n",
              "      <td>3.6</td>\n",
              "      <td>2.7</td>\n",
              "      <td>2.7</td>\n",
              "      <td>2.7</td>\n",
              "      <td>12.7</td>\n",
              "      <td>12.8</td>\n",
              "      <td>1461</td>\n",
              "      <td>25</td>\n",
              "      <td>72.2</td>\n",
              "      <td>2.2</td>\n",
              "      <td>15.5</td>\n",
              "      <td>3.6</td>\n",
              "      <td>6.5</td>\n",
              "      <td>25.4</td>\n",
              "      <td>0.77</td>\n",
              "      <td>95.4</td>\n",
              "      <td>2.8</td>\n",
              "      <td>8.8</td>\n",
              "      <td>15.8</td>\n",
              "      <td>30.1</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>114065</th>\n",
              "      <td>877336</td>\n",
              "      <td>2020-07-25 11:52:58</td>\n",
              "      <td>Public Payphone Complaint</td>\n",
              "      <td>DOITT</td>\n",
              "      <td>3 QUEENS</td>\n",
              "      <td>40.751321</td>\n",
              "      <td>-73.862459</td>\n",
              "      <td>graffiti litter on phone</td>\n",
              "      <td>QUEENS</td>\n",
              "      <td>2020-07-25</td>\n",
              "      <td>7</td>\n",
              "      <td>5</td>\n",
              "      <td>11</td>\n",
              "      <td>10</td>\n",
              "      <td>4</td>\n",
              "      <td>3</td>\n",
              "      <td>13</td>\n",
              "      <td>8.4</td>\n",
              "      <td>3.0</td>\n",
              "      <td>3.9</td>\n",
              "      <td>4.2</td>\n",
              "      <td>3.9</td>\n",
              "      <td>3.4</td>\n",
              "      <td>3.2</td>\n",
              "      <td>2.9</td>\n",
              "      <td>2.7</td>\n",
              "      <td>2.5</td>\n",
              "      <td>2.5</td>\n",
              "      <td>3.1</td>\n",
              "      <td>4.8</td>\n",
              "      <td>4.4</td>\n",
              "      <td>3.9</td>\n",
              "      <td>3.4</td>\n",
              "      <td>3.1</td>\n",
              "      <td>1.9</td>\n",
              "      <td>22.2</td>\n",
              "      <td>12.7</td>\n",
              "      <td>14658</td>\n",
              "      <td>225</td>\n",
              "      <td>10.4</td>\n",
              "      <td>5.0</td>\n",
              "      <td>16.5</td>\n",
              "      <td>1.5</td>\n",
              "      <td>66.6</td>\n",
              "      <td>42.9</td>\n",
              "      <td>0.01</td>\n",
              "      <td>94.3</td>\n",
              "      <td>3.5</td>\n",
              "      <td>24.0</td>\n",
              "      <td>28.8</td>\n",
              "      <td>53.4</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>114066</th>\n",
              "      <td>171680</td>\n",
              "      <td>2020-11-17 16:49:12</td>\n",
              "      <td>Linknyc</td>\n",
              "      <td>DOITT</td>\n",
              "      <td>4 MANHATTAN</td>\n",
              "      <td>40.743605</td>\n",
              "      <td>-73.999726</td>\n",
              "      <td>damaged defective</td>\n",
              "      <td>MANHATTAN</td>\n",
              "      <td>2020-11-17</td>\n",
              "      <td>11</td>\n",
              "      <td>1</td>\n",
              "      <td>16</td>\n",
              "      <td>38</td>\n",
              "      <td>2</td>\n",
              "      <td>7</td>\n",
              "      <td>29</td>\n",
              "      <td>18.1</td>\n",
              "      <td>4.0</td>\n",
              "      <td>7.2</td>\n",
              "      <td>6.4</td>\n",
              "      <td>3.8</td>\n",
              "      <td>3.0</td>\n",
              "      <td>2.9</td>\n",
              "      <td>2.3</td>\n",
              "      <td>2.7</td>\n",
              "      <td>2.7</td>\n",
              "      <td>1.1</td>\n",
              "      <td>3.3</td>\n",
              "      <td>7.2</td>\n",
              "      <td>4.8</td>\n",
              "      <td>3.9</td>\n",
              "      <td>3.6</td>\n",
              "      <td>3.7</td>\n",
              "      <td>3.1</td>\n",
              "      <td>8.2</td>\n",
              "      <td>13.4</td>\n",
              "      <td>3417</td>\n",
              "      <td>70</td>\n",
              "      <td>59.1</td>\n",
              "      <td>5.2</td>\n",
              "      <td>17.2</td>\n",
              "      <td>3.0</td>\n",
              "      <td>15.5</td>\n",
              "      <td>26.7</td>\n",
              "      <td>0.18</td>\n",
              "      <td>95.6</td>\n",
              "      <td>3.5</td>\n",
              "      <td>11.3</td>\n",
              "      <td>15.8</td>\n",
              "      <td>33.9</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>114067</th>\n",
              "      <td>306584</td>\n",
              "      <td>2020-06-18 13:35:15</td>\n",
              "      <td>Public Payphone Complaint</td>\n",
              "      <td>DOITT</td>\n",
              "      <td>5 MANHATTAN</td>\n",
              "      <td>40.763025</td>\n",
              "      <td>-73.978139</td>\n",
              "      <td>no dial tone</td>\n",
              "      <td>MANHATTAN</td>\n",
              "      <td>2020-06-18</td>\n",
              "      <td>6</td>\n",
              "      <td>3</td>\n",
              "      <td>13</td>\n",
              "      <td>23</td>\n",
              "      <td>7</td>\n",
              "      <td>3</td>\n",
              "      <td>13</td>\n",
              "      <td>29.4</td>\n",
              "      <td>4.0</td>\n",
              "      <td>7.2</td>\n",
              "      <td>6.4</td>\n",
              "      <td>3.8</td>\n",
              "      <td>3.0</td>\n",
              "      <td>2.9</td>\n",
              "      <td>2.3</td>\n",
              "      <td>2.7</td>\n",
              "      <td>2.7</td>\n",
              "      <td>1.1</td>\n",
              "      <td>3.3</td>\n",
              "      <td>7.2</td>\n",
              "      <td>4.8</td>\n",
              "      <td>3.9</td>\n",
              "      <td>3.6</td>\n",
              "      <td>3.7</td>\n",
              "      <td>3.1</td>\n",
              "      <td>8.2</td>\n",
              "      <td>13.4</td>\n",
              "      <td>3018</td>\n",
              "      <td>109</td>\n",
              "      <td>59.1</td>\n",
              "      <td>5.2</td>\n",
              "      <td>17.2</td>\n",
              "      <td>3.0</td>\n",
              "      <td>15.5</td>\n",
              "      <td>26.7</td>\n",
              "      <td>0.17</td>\n",
              "      <td>94.4</td>\n",
              "      <td>3.5</td>\n",
              "      <td>11.3</td>\n",
              "      <td>15.8</td>\n",
              "      <td>33.9</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>114068</th>\n",
              "      <td>459783</td>\n",
              "      <td>2020-08-17 09:46:23</td>\n",
              "      <td>Public Payphone Complaint</td>\n",
              "      <td>DOITT</td>\n",
              "      <td>3 MANHATTAN</td>\n",
              "      <td>40.717612</td>\n",
              "      <td>-73.992035</td>\n",
              "      <td>graffiti litter on phone</td>\n",
              "      <td>MANHATTAN</td>\n",
              "      <td>2020-08-17</td>\n",
              "      <td>8</td>\n",
              "      <td>0</td>\n",
              "      <td>9</td>\n",
              "      <td>41</td>\n",
              "      <td>5</td>\n",
              "      <td>13</td>\n",
              "      <td>42</td>\n",
              "      <td>15.2</td>\n",
              "      <td>4.2</td>\n",
              "      <td>6.4</td>\n",
              "      <td>4.1</td>\n",
              "      <td>3.2</td>\n",
              "      <td>3.0</td>\n",
              "      <td>3.2</td>\n",
              "      <td>2.8</td>\n",
              "      <td>3.3</td>\n",
              "      <td>3.4</td>\n",
              "      <td>2.5</td>\n",
              "      <td>3.7</td>\n",
              "      <td>5.6</td>\n",
              "      <td>2.9</td>\n",
              "      <td>3.3</td>\n",
              "      <td>3.0</td>\n",
              "      <td>2.8</td>\n",
              "      <td>2.5</td>\n",
              "      <td>11.4</td>\n",
              "      <td>17.9</td>\n",
              "      <td>4259</td>\n",
              "      <td>109</td>\n",
              "      <td>33.3</td>\n",
              "      <td>8.4</td>\n",
              "      <td>30.6</td>\n",
              "      <td>2.6</td>\n",
              "      <td>25.1</td>\n",
              "      <td>31.3</td>\n",
              "      <td>-0.01</td>\n",
              "      <td>91.0</td>\n",
              "      <td>3.6</td>\n",
              "      <td>19.3</td>\n",
              "      <td>15.8</td>\n",
              "      <td>38.4</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>\n",
              "<p>114069 rows × 51 columns</p>\n",
              "</div>"
            ],
            "text/plain": [
              "          index         created_date  ... lep_rate_boro pct_hh_rent_burd\n",
              "0        383700  2020-10-02 10:07:39  ...          23.6             48.1\n",
              "1        164754  2020-09-15 11:47:20  ...          15.8             33.9\n",
              "2       1301458  2020-07-02 15:02:08  ...          23.6             28.8\n",
              "3        661127  2020-08-29 15:41:01  ...          28.8             46.9\n",
              "4        252987  2020-06-04 13:20:26  ...          23.6             49.9\n",
              "...         ...                  ...  ...           ...              ...\n",
              "114064  1003068  2020-09-29 14:16:21  ...          15.8             30.1\n",
              "114065   877336  2020-07-25 11:52:58  ...          28.8             53.4\n",
              "114066   171680  2020-11-17 16:49:12  ...          15.8             33.9\n",
              "114067   306584  2020-06-18 13:35:15  ...          15.8             33.9\n",
              "114068   459783  2020-08-17 09:46:23  ...          15.8             38.4\n",
              "\n",
              "[114069 rows x 51 columns]"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 204
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "VJ-R1hGCuT4P"
      },
      "source": [
        "In order for the model to be tested on new data, it should have exposure to every unique descriptor in the dataset. The below code creates a list of every item in the original dataframe's unique descriptors that is not found in the sample dataframe."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "LCKQkzBWcAJy",
        "outputId": "7c038c27-ae3c-47d5-9406-ccdc245f5fef"
      },
      "source": [
        "descriptors_not_in_sample = []\n",
        "\n",
        "#update above list with descriptors that aren't in the sample \n",
        "for i in df.descriptor.unique():\n",
        "  if i not in weighted_df.descriptor.unique():\n",
        "    descriptors_not_in_sample.append(i)\n",
        "\n",
        "len(descriptors_not_in_sample)"
      ],
      "execution_count": 205,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "168"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 205
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "Iim78erDu8if"
      },
      "source": [
        "The list above captures many of the less common descriptors that were not captured in the sample data, including most descriptors associated with minority agencies. Below, we create a dataframe with all calls that match these descriptors and concatenate them with the weighted sample dataframe."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "zdZUiYEveDap"
      },
      "source": [
        "descriptors_not_in_sample_df = df[df.descriptor.isin(descriptors_not_in_sample)]"
      ],
      "execution_count": 206,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "XTy5NgSAezLu"
      },
      "source": [
        "weighted_df = pd.concat([weighted_df,\n",
        "                         descriptors_not_in_sample_df])\n",
        "weighted_df.reset_index(inplace=True)"
      ],
      "execution_count": 207,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "Ka9F8jA5b1xJ",
        "outputId": "39193375-9f70-476e-92b1-96fcbe33e8f7"
      },
      "source": [
        "#confirm that the number of unique descriptors is the same for both dataframes\n",
        "len(weighted_df.descriptor.unique()) == len(df.descriptor.unique())"
      ],
      "execution_count": 208,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "True"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 208
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "7G-L7vYfTWXO"
      },
      "source": [
        "The classes remain imbalanced, but no agency is represented by only a handful of samples, as would have been the case with simple random sampling:"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "iJ_kfxRaDU69",
        "outputId": "6eebc8b2-d7c7-4df7-e78a-5b784b874cce"
      },
      "source": [
        "weighted_df.agency.value_counts()"
      ],
      "execution_count": 19,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "NYPD     60042\n",
              "HPD      14521\n",
              "DPR       9556\n",
              "DOT       6999\n",
              "DEP       6684\n",
              "DSNY      5022\n",
              "OSE       3787\n",
              "DOB       3618\n",
              "DOHMH     2675\n",
              "DCA        978\n",
              "EDC        474\n",
              "TLC        412\n",
              "DOE         53\n",
              "DOITT       37\n",
              "Name: agency, dtype: int64"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 19
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "0Xxzm3adUPw4"
      },
      "source": [
        "# Keras Neural Network"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "mEdxA7s6wc6X"
      },
      "source": [
        "## Preparing the Training and Test Data"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "rQ45Wmxxwhg0"
      },
      "source": [
        "Using the dataframe of weighted samples, we assign 'X' to the descriptor column and 'y' to the agency column. "
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "XuDY4PxXbCXy"
      },
      "source": [
        "X = weighted_df.descriptor\n",
        "y = weighted_df.agency"
      ],
      "execution_count": 209,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "vWtSr0A5w3qD"
      },
      "source": [
        "The [Keras Tokenizer](https://www.tensorflow.org/api_docs/python/tf/keras/preprocessing/text/Tokenizer) `fit_on_texts()` method, used below, \"updates internal vocabulary based on a list of texts.\" In this case, the list of texts is all descriptors in the weighted sample. The `texts_to_matrix()` method converts the list of texts to a NumPy matrix. Specifying the mode \"binary\" means that only the presence or absence of each item in the tokenizer's vocabulary will be accounted for in the matrix."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "0q7gzxcYgyQ2"
      },
      "source": [
        "descriptors = X\n",
        "tokenizer = Tokenizer() \n",
        "\n",
        "tokenizer.fit_on_texts(descriptors) \n",
        "descriptor_onehot = tokenizer.texts_to_matrix(descriptors, mode='binary')\n",
        "word_index = tokenizer.word_index"
      ],
      "execution_count": 210,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "yQMyujM_-goW"
      },
      "source": [
        "The word index is a dictionary that contains words in the tokenizer's vocabulary as keys, and integers that correspond to the keys' frequency rank as values. The first ten items of the dictionary are printed below, and we can see that the most frequent words have the lowest values. The highest value in the dictionary is 1114."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "tqSVnoIz3KfY",
        "outputId": "6d5fafd8-8bc3-44d3-a18d-e4090937fd0f"
      },
      "source": [
        "print(list(word_index.items())[0:10])\n",
        "print('Max:', max(word_index.values()))"
      ],
      "execution_count": 211,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "[('music', 1), ('loud', 2), ('party', 3), ('access', 4), ('no', 5), ('blocked', 6), ('hydrant', 7), ('truck', 8), ('car', 9), ('in', 10)]\n",
            "Max: 1114\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "58-ghEfZB5Hn"
      },
      "source": [
        "The content and shape of the one-hot encoded descriptors are displayed below."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 158
        },
        "id": "W1L7LoTr3fA8",
        "outputId": "3fe9f0e2-5177-41ff-8b54-1fe87dc52e5d"
      },
      "source": [
        "display(descriptor_onehot)\n",
        "\n",
        "print('Shape:', descriptor_onehot.shape)"
      ],
      "execution_count": 212,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "array([[0., 0., 0., ..., 0., 0., 0.],\n",
              "       [0., 0., 0., ..., 0., 0., 0.],\n",
              "       [0., 0., 0., ..., 0., 0., 0.],\n",
              "       ...,\n",
              "       [0., 0., 0., ..., 0., 0., 0.],\n",
              "       [0., 0., 0., ..., 0., 0., 0.],\n",
              "       [0., 0., 0., ..., 0., 0., 0.]])"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "stream",
          "text": [
            "Shape: (114858, 1115)\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "NWO_IAYWC4Mc"
      },
      "source": [
        "The number of rows is the same as the number of rows in X, and correspond to an individual call:"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "1zOwlHkJCtrZ",
        "outputId": "12671727-bae7-44e4-b226-081abda5b7fd"
      },
      "source": [
        "X.shape"
      ],
      "execution_count": 213,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "(114858,)"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 213
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "KEr1U7bzDknt"
      },
      "source": [
        "Each column corresponds to one of 1,110 words in the tokenizer's vocabulary. If the word for a column is present in a given call's descriptor, its value at the call's corresponding row will be '1.' Otherwise, it will be '0.'"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "EIw59i2XFJ12"
      },
      "source": [
        "The label encoder used below assigns a unique integer to each of the categories in the agency column. The [keras.utils.to_categorical\n",
        "](https://www.tensorflow.org/api_docs/python/tf/keras/utils/to_categorical) method \"converts a class vector (integers) to binary class matrix.\""
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "cJGOuSOohgMX"
      },
      "source": [
        "agency = y\n",
        "\n",
        "le = preprocessing.LabelEncoder() \n",
        "le.fit(agency)\n",
        "agency_cat = le.transform(agency)  \n",
        "agency_onehot = to_categorical(agency_cat)"
      ],
      "execution_count": 214,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "BitB8V1fKuFb"
      },
      "source": [
        "Let's examine the content and shape of the encoded agency data:"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 158
        },
        "id": "XGhGFhjcKYHQ",
        "outputId": "c19e9fe2-42a4-4210-9d33-01430d955141"
      },
      "source": [
        "display(agency_onehot)\n",
        "print('Shape:', agency_onehot.shape)"
      ],
      "execution_count": 26,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "array([[0., 0., 1., ..., 0., 0., 0.],\n",
              "       [0., 0., 1., ..., 0., 0., 0.],\n",
              "       [0., 0., 1., ..., 0., 0., 0.],\n",
              "       ...,\n",
              "       [0., 0., 0., ..., 0., 0., 0.],\n",
              "       [0., 0., 0., ..., 0., 0., 0.],\n",
              "       [0., 0., 0., ..., 0., 0., 0.]], dtype=float32)"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "stream",
          "text": [
            "Shape: (114858, 14)\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "fe7mqmR7QA6y"
      },
      "source": [
        "Each row corresponds to a call, and each column corresponds to an agency. The value in a given cell indicates whether the agency in its column either responded ('1'), or did not respond ('0'), to the call in its row.\n",
        "\n",
        "Now that both X and y have been encoded, we'll create a random split of training and test data."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "w17CTxizTvsS"
      },
      "source": [
        "test, train, label_test, label_train = train_test_split(descriptor_onehot, agency_onehot, test_size=.1, random_state=6)"
      ],
      "execution_count": 215,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "fcydqkeWxrQt"
      },
      "source": [
        "## Baseline Model"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "fQ7LdKjaI-kH"
      },
      "source": [
        "We'll start with a baseline model before writing a function to assess how different parameters affect the model's performance. For all of the models, we'll use the \"categorical crossentropy\" loss function, which is appropriate for multiclass classification. Accuracy will also be the performance metric used for each model, but our interpretation of the accuracy should be informed by the large number of classes and significant class imbalance."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "xkP7CC257B1v"
      },
      "source": [
        "model1 = models.Sequential()\n",
        "\n",
        "model1.add(layers.Dense(50, activation='relu', input_shape=(test.shape[1],)))\n",
        "model1.add(layers.Dense(25, activation='relu'))\n",
        "model1.add(layers.Dense(14, activation='softmax'))\n",
        "\n",
        "\n",
        "opt = tf.keras.optimizers.SGD(learning_rate=0.001)\n",
        "model1.compile(optimizer=opt,\n",
        "              loss='categorical_crossentropy',\n",
        "              metrics=['acc'])"
      ],
      "execution_count": 28,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "889WTz47R5_A"
      },
      "source": [
        "To avoid overfitting the model to the training data, we'll add the `validation_split` parameter, and an early stopping callback that monitors validation loss, to our fit function."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "cRgLsdFd7enu",
        "outputId": "f26191e0-bce1-42d2-acbe-6fe7efd75e52"
      },
      "source": [
        "es = keras.callbacks.EarlyStopping(monitor='val_loss', patience=5)\n",
        "\n",
        "history = model1.fit(train,\n",
        "                    label_train,\n",
        "                    validation_split=0.20,\n",
        "                    epochs=62,\n",
        "                    batch_size=32,\n",
        "                    callbacks=[es])"
      ],
      "execution_count": 29,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Epoch 1/62\n",
            "288/288 [==============================] - 1s 3ms/step - loss: 2.5534 - acc: 0.4877 - val_loss: 2.4941 - val_acc: 0.5109\n",
            "Epoch 2/62\n",
            "288/288 [==============================] - 1s 2ms/step - loss: 2.4278 - acc: 0.5215 - val_loss: 2.3713 - val_acc: 0.5109\n",
            "Epoch 3/62\n",
            "288/288 [==============================] - 1s 2ms/step - loss: 2.3006 - acc: 0.5215 - val_loss: 2.2430 - val_acc: 0.5109\n",
            "Epoch 4/62\n",
            "288/288 [==============================] - 1s 2ms/step - loss: 2.1633 - acc: 0.5215 - val_loss: 2.1039 - val_acc: 0.5109\n",
            "Epoch 5/62\n",
            "288/288 [==============================] - 1s 2ms/step - loss: 2.0177 - acc: 0.5215 - val_loss: 1.9634 - val_acc: 0.5109\n",
            "Epoch 6/62\n",
            "288/288 [==============================] - 1s 2ms/step - loss: 1.8843 - acc: 0.5215 - val_loss: 1.8488 - val_acc: 0.5109\n",
            "Epoch 7/62\n",
            "288/288 [==============================] - 1s 2ms/step - loss: 1.7839 - acc: 0.5215 - val_loss: 1.7688 - val_acc: 0.5109\n",
            "Epoch 8/62\n",
            "288/288 [==============================] - 1s 2ms/step - loss: 1.7156 - acc: 0.5215 - val_loss: 1.7152 - val_acc: 0.5109\n",
            "Epoch 9/62\n",
            "288/288 [==============================] - 1s 2ms/step - loss: 1.6682 - acc: 0.5215 - val_loss: 1.6755 - val_acc: 0.5109\n",
            "Epoch 10/62\n",
            "288/288 [==============================] - 1s 2ms/step - loss: 1.6313 - acc: 0.5215 - val_loss: 1.6427 - val_acc: 0.5109\n",
            "Epoch 11/62\n",
            "288/288 [==============================] - 1s 2ms/step - loss: 1.6001 - acc: 0.5215 - val_loss: 1.6140 - val_acc: 0.5109\n",
            "Epoch 12/62\n",
            "288/288 [==============================] - 1s 2ms/step - loss: 1.5723 - acc: 0.5215 - val_loss: 1.5881 - val_acc: 0.5109\n",
            "Epoch 13/62\n",
            "288/288 [==============================] - 1s 2ms/step - loss: 1.5471 - acc: 0.5215 - val_loss: 1.5644 - val_acc: 0.5109\n",
            "Epoch 14/62\n",
            "288/288 [==============================] - 1s 2ms/step - loss: 1.5238 - acc: 0.5215 - val_loss: 1.5423 - val_acc: 0.5109\n",
            "Epoch 15/62\n",
            "288/288 [==============================] - 1s 2ms/step - loss: 1.5020 - acc: 0.5215 - val_loss: 1.5213 - val_acc: 0.5109\n",
            "Epoch 16/62\n",
            "288/288 [==============================] - 1s 2ms/step - loss: 1.4812 - acc: 0.5215 - val_loss: 1.5013 - val_acc: 0.5109\n",
            "Epoch 17/62\n",
            "288/288 [==============================] - 1s 2ms/step - loss: 1.4614 - acc: 0.5215 - val_loss: 1.4821 - val_acc: 0.5109\n",
            "Epoch 18/62\n",
            "288/288 [==============================] - 1s 2ms/step - loss: 1.4424 - acc: 0.5215 - val_loss: 1.4637 - val_acc: 0.5109\n",
            "Epoch 19/62\n",
            "288/288 [==============================] - 1s 2ms/step - loss: 1.4242 - acc: 0.5215 - val_loss: 1.4459 - val_acc: 0.5109\n",
            "Epoch 20/62\n",
            "288/288 [==============================] - 1s 2ms/step - loss: 1.4065 - acc: 0.5215 - val_loss: 1.4286 - val_acc: 0.5109\n",
            "Epoch 21/62\n",
            "288/288 [==============================] - 1s 2ms/step - loss: 1.3892 - acc: 0.5215 - val_loss: 1.4116 - val_acc: 0.5109\n",
            "Epoch 22/62\n",
            "288/288 [==============================] - 1s 2ms/step - loss: 1.3726 - acc: 0.5215 - val_loss: 1.3953 - val_acc: 0.5109\n",
            "Epoch 23/62\n",
            "288/288 [==============================] - 1s 2ms/step - loss: 1.3566 - acc: 0.5215 - val_loss: 1.3795 - val_acc: 0.5109\n",
            "Epoch 24/62\n",
            "288/288 [==============================] - 1s 2ms/step - loss: 1.3410 - acc: 0.5215 - val_loss: 1.3643 - val_acc: 0.5109\n",
            "Epoch 25/62\n",
            "288/288 [==============================] - 1s 2ms/step - loss: 1.3261 - acc: 0.5215 - val_loss: 1.3496 - val_acc: 0.5109\n",
            "Epoch 26/62\n",
            "288/288 [==============================] - 1s 2ms/step - loss: 1.3116 - acc: 0.5215 - val_loss: 1.3354 - val_acc: 0.5109\n",
            "Epoch 27/62\n",
            "288/288 [==============================] - 1s 2ms/step - loss: 1.2975 - acc: 0.5215 - val_loss: 1.3216 - val_acc: 0.5109\n",
            "Epoch 28/62\n",
            "288/288 [==============================] - 1s 2ms/step - loss: 1.2840 - acc: 0.5247 - val_loss: 1.3083 - val_acc: 0.5335\n",
            "Epoch 29/62\n",
            "288/288 [==============================] - 1s 2ms/step - loss: 1.2710 - acc: 0.5409 - val_loss: 1.2955 - val_acc: 0.5335\n",
            "Epoch 30/62\n",
            "288/288 [==============================] - 1s 2ms/step - loss: 1.2583 - acc: 0.5409 - val_loss: 1.2831 - val_acc: 0.5335\n",
            "Epoch 31/62\n",
            "288/288 [==============================] - 1s 2ms/step - loss: 1.2461 - acc: 0.5409 - val_loss: 1.2710 - val_acc: 0.5335\n",
            "Epoch 32/62\n",
            "288/288 [==============================] - 1s 2ms/step - loss: 1.2342 - acc: 0.5489 - val_loss: 1.2593 - val_acc: 0.5492\n",
            "Epoch 33/62\n",
            "288/288 [==============================] - 1s 2ms/step - loss: 1.2227 - acc: 0.5595 - val_loss: 1.2479 - val_acc: 0.5500\n",
            "Epoch 34/62\n",
            "288/288 [==============================] - 1s 2ms/step - loss: 1.2115 - acc: 0.5697 - val_loss: 1.2369 - val_acc: 0.5714\n",
            "Epoch 35/62\n",
            "288/288 [==============================] - 1s 2ms/step - loss: 1.2007 - acc: 0.5902 - val_loss: 1.2262 - val_acc: 0.5788\n",
            "Epoch 36/62\n",
            "288/288 [==============================] - 1s 2ms/step - loss: 1.1902 - acc: 0.5945 - val_loss: 1.2158 - val_acc: 0.5783\n",
            "Epoch 37/62\n",
            "288/288 [==============================] - 1s 2ms/step - loss: 1.1800 - acc: 0.5963 - val_loss: 1.2058 - val_acc: 0.5818\n",
            "Epoch 38/62\n",
            "288/288 [==============================] - 1s 2ms/step - loss: 1.1701 - acc: 0.6075 - val_loss: 1.1960 - val_acc: 0.6118\n",
            "Epoch 39/62\n",
            "288/288 [==============================] - 1s 2ms/step - loss: 1.1605 - acc: 0.6293 - val_loss: 1.1865 - val_acc: 0.6158\n",
            "Epoch 40/62\n",
            "288/288 [==============================] - 1s 2ms/step - loss: 1.1512 - acc: 0.6331 - val_loss: 1.1772 - val_acc: 0.6232\n",
            "Epoch 41/62\n",
            "288/288 [==============================] - 1s 2ms/step - loss: 1.1421 - acc: 0.6357 - val_loss: 1.1682 - val_acc: 0.6232\n",
            "Epoch 42/62\n",
            "288/288 [==============================] - 1s 2ms/step - loss: 1.1333 - acc: 0.6369 - val_loss: 1.1595 - val_acc: 0.6232\n",
            "Epoch 43/62\n",
            "288/288 [==============================] - 1s 2ms/step - loss: 1.1247 - acc: 0.6369 - val_loss: 1.1509 - val_acc: 0.6214\n",
            "Epoch 44/62\n",
            "288/288 [==============================] - 1s 2ms/step - loss: 1.1163 - acc: 0.6391 - val_loss: 1.1425 - val_acc: 0.6236\n",
            "Epoch 45/62\n",
            "288/288 [==============================] - 1s 2ms/step - loss: 1.1081 - acc: 0.6397 - val_loss: 1.1344 - val_acc: 0.6271\n",
            "Epoch 46/62\n",
            "288/288 [==============================] - 1s 2ms/step - loss: 1.1001 - acc: 0.6421 - val_loss: 1.1264 - val_acc: 0.6284\n",
            "Epoch 47/62\n",
            "288/288 [==============================] - 1s 2ms/step - loss: 1.0923 - acc: 0.6420 - val_loss: 1.1185 - val_acc: 0.6292\n",
            "Epoch 48/62\n",
            "288/288 [==============================] - 1s 2ms/step - loss: 1.0847 - acc: 0.6419 - val_loss: 1.1108 - val_acc: 0.6275\n",
            "Epoch 49/62\n",
            "288/288 [==============================] - 1s 2ms/step - loss: 1.0772 - acc: 0.6412 - val_loss: 1.1032 - val_acc: 0.6275\n",
            "Epoch 50/62\n",
            "288/288 [==============================] - 1s 2ms/step - loss: 1.0698 - acc: 0.6515 - val_loss: 1.0958 - val_acc: 0.6449\n",
            "Epoch 51/62\n",
            "288/288 [==============================] - 1s 2ms/step - loss: 1.0625 - acc: 0.6579 - val_loss: 1.0884 - val_acc: 0.6445\n",
            "Epoch 52/62\n",
            "288/288 [==============================] - 1s 2ms/step - loss: 1.0553 - acc: 0.6574 - val_loss: 1.0811 - val_acc: 0.6445\n",
            "Epoch 53/62\n",
            "288/288 [==============================] - 1s 2ms/step - loss: 1.0482 - acc: 0.6574 - val_loss: 1.0739 - val_acc: 0.6445\n",
            "Epoch 54/62\n",
            "288/288 [==============================] - 1s 2ms/step - loss: 1.0413 - acc: 0.6573 - val_loss: 1.0668 - val_acc: 0.6445\n",
            "Epoch 55/62\n",
            "288/288 [==============================] - 1s 3ms/step - loss: 1.0345 - acc: 0.6573 - val_loss: 1.0597 - val_acc: 0.6445\n",
            "Epoch 56/62\n",
            "288/288 [==============================] - 1s 2ms/step - loss: 1.0276 - acc: 0.6573 - val_loss: 1.0527 - val_acc: 0.6445\n",
            "Epoch 57/62\n",
            "288/288 [==============================] - 1s 2ms/step - loss: 1.0208 - acc: 0.6573 - val_loss: 1.0457 - val_acc: 0.6445\n",
            "Epoch 58/62\n",
            "288/288 [==============================] - 1s 2ms/step - loss: 1.0140 - acc: 0.6571 - val_loss: 1.0386 - val_acc: 0.6445\n",
            "Epoch 59/62\n",
            "288/288 [==============================] - 1s 2ms/step - loss: 1.0072 - acc: 0.6572 - val_loss: 1.0316 - val_acc: 0.6475\n",
            "Epoch 60/62\n",
            "288/288 [==============================] - 1s 2ms/step - loss: 1.0005 - acc: 0.6620 - val_loss: 1.0247 - val_acc: 0.6475\n",
            "Epoch 61/62\n",
            "288/288 [==============================] - 1s 2ms/step - loss: 0.9938 - acc: 0.6697 - val_loss: 1.0178 - val_acc: 0.6636\n",
            "Epoch 62/62\n",
            "288/288 [==============================] - 1s 2ms/step - loss: 0.9871 - acc: 0.6772 - val_loss: 1.0109 - val_acc: 0.6636\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "HXCkp3-l0FRD"
      },
      "source": [
        "Let's look at the model's loss and accuracy score."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "Fpz0BAQqKNf9",
        "outputId": "da8fa0e9-970d-45c2-bfbc-84e76c2623b5"
      },
      "source": [
        "results = model1.evaluate(test, label_test, batch_size=32)"
      ],
      "execution_count": 30,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "3231/3231 [==============================] - 5s 2ms/step - loss: 0.9842 - acc: 0.6742\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "KRKe8u-B0Nu5"
      },
      "source": [
        "The function below plots the model's loss and accuracy curves, which will provide a better visual sense of how the values for these metrics changed as the number of epochs increased."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "amYF3K8pbej4"
      },
      "source": [
        "def plot_loss_acc(history):\n",
        " \n",
        "  \"\"\"Takes in a Keras history object and returns a 15*6 figure displaying the \n",
        "     model's loss and accuracy curves.\"\"\"\n",
        "\n",
        "\n",
        "  fig, (ax1, ax2) = plt.subplots(nrows=1,\n",
        "                               ncols=2,\n",
        "                               figsize=(15, 6))\n",
        "\n",
        "  ax_font = {'fontsize': 12, 'family': 'serif', \n",
        "             'color': '#58508d', 'weight': 'bold'}\n",
        "  title_font = {'fontsize': 14, 'family': 'serif',\n",
        "                'color': '#58508d', 'weight': 'bold'}\n",
        "\n",
        "\n",
        "  ax1.plot(history.history['acc'], color='#D8A7B1', linewidth=3)\n",
        "  ax1.grid(color='g', linestyle='-', linewidth=.09)\n",
        "  ax1.plot(history.history['val_acc'], color='#05234C', linewidth=3)\n",
        "  ax1.set_title('Model Accuracy', fontdict=title_font)\n",
        "  ax1.set_ylabel('Accuracy', fontdict=ax_font)\n",
        "  ax1.set_xlabel('Number of epochs', fontdict=ax_font)\n",
        "  ax1.legend(['train', 'validation'], loc='upper left')\n",
        "\n",
        "\n",
        "  ax2.plot(history.history['loss'], color='#D8A7B1', linewidth=3)\n",
        "  ax2.grid(color='g', linestyle='-', linewidth=.09)\n",
        "  ax2.plot(history.history['val_loss'], color='#05234C', linewidth=3)\n",
        "  ax2.set_title('Model Loss', fontdict=title_font)\n",
        "  ax2.set_ylabel('Loss', fontdict=ax_font)\n",
        "  ax2.set_xlabel('Number of epochs', fontdict=ax_font)\n",
        "  ax2.legend(['train', 'validation'], loc='upper right')\n",
        "\n",
        "  fig.show()"
      ],
      "execution_count": 31,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 407
        },
        "id": "UWoFuBe5hLco",
        "outputId": "9aa1ca45-cb39-455e-ce38-92fbc9625479"
      },
      "source": [
        "plot_loss_acc(history)"
      ],
      "execution_count": 32,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 1080x432 with 2 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "n4iIKXNU0qcn"
      },
      "source": [
        "The accuracy score improves steadily after 30 epochs, while the loss steadily drops. We'll also create a random subset of the data that has not been resampled to evaluate the model's performance, and apply the same preprocessing steps as above."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "ywhDpnurahU8"
      },
      "source": [
        "random_df = df.sample(len(weighted_df), random_state=6)"
      ],
      "execution_count": 33,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "yuQ8NFhstPu8",
        "outputId": "ba9a1eb9-0496-448d-ddd7-040956630816"
      },
      "source": [
        "descriptors_not_in_sample = []\n",
        "\n",
        "for i in df.descriptor.unique():\n",
        "  if i not in random_df.descriptor.unique():\n",
        "    descriptors_not_in_sample.append(i)\n",
        "\n",
        "descriptors_not_in_sample"
      ],
      "execution_count": 34,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "['bingo hall',\n",
              " 'waterwaycolor ihd',\n",
              " 'oil, grease in water qd',\n",
              " 'safety inspectionretaining walls may ',\n",
              " 'allergy information',\n",
              " 'other industrial waste problem use comments for dep internal use only iz',\n",
              " 'indoor pool over capacity',\n",
              " 'jewelry appraiser',\n",
              " 'graffiti litter on phone',\n",
              " 'phone blocking sidewalk',\n",
              " 'sediment',\n",
              " 'snake',\n",
              " 'high grass',\n",
              " 'cloudy or milky, other use comments qbz',\n",
              " 'push button',\n",
              " 'cloudy or milky water qb',\n",
              " 'absent',\n",
              " 'inspection work order dob',\n",
              " 'cigarette vending machine',\n",
              " 'mast arm',\n",
              " 'unauthorized posting of signs',\n",
              " 'ped flasher',\n",
              " 'enclosure cap',\n",
              " 'maintenance cover',\n",
              " 'booting company',\n",
              " 'warning signal lamp',\n",
              " 'immigration services',\n",
              " 'oil spill into basin sewer  small iabs',\n",
              " 'roof gutters',\n",
              " 'platform condition',\n",
              " 'manhole cover broken making noise sb',\n",
              " 'dogs or cats not sold',\n",
              " 'bent loose',\n",
              " 'toxic chemical material',\n",
              " 'basement',\n",
              " 'gasoline spill iaa',\n",
              " 'puddle on sidewalk',\n",
              " 'warning signal',\n",
              " 'remove debris',\n",
              " 'plants odor related problems po',\n",
              " 'street con game',\n",
              " 'temporary',\n",
              " 'inadequate support shoring',\n",
              " 'building',\n",
              " 'other water problem use comments qzz',\n",
              " 'shisha',\n",
              " 'used gun dealer',\n",
              " 'natural gas leak *for dep internal use only* hf',\n",
              " 'junction box',\n",
              " 'veh signal lens',\n",
              " 'grocery store',\n",
              " 'plate missing movedexposing hole sb',\n",
              " 'e transfer station',\n",
              " 'defective water sampling station qss',\n",
              " 'use of newly seeded lawn',\n",
              " 'color',\n",
              " 'air: odor fumes, dry cleaners ad',\n",
              " 'disabled device dealer',\n",
              " 'smoking ban  smoking on construction site',\n",
              " 'new automatic public toilet request',\n",
              " 'driver complaint',\n",
              " 'musical instrument',\n",
              " 'sports equipment',\n",
              " ' litter basket   request',\n",
              " 'contamination risk',\n",
              " ' dead animals',\n",
              " 'inattentive',\n",
              " 'user unlicensed',\n",
              " 'amusement ride',\n",
              " 'integrity complaint referral',\n",
              " 'curb violation',\n",
              " 'flood light lamp missing',\n",
              " 'plumbing problem',\n",
              " 'sway bar',\n",
              " 'on messenger',\n",
              " 'b sweeping inadequate',\n",
              " 'eb sidewalk obstruction',\n",
              " 'general bad condition',\n",
              " 'relocation of parking meter',\n",
              " 'ped lens',\n",
              " 'guard rail  bridge',\n",
              " 'catch basin search sc',\n",
              " 'c uncollected xmas trees',\n",
              " 'controller flasher',\n",
              " 'foundation',\n",
              " 'request to open a hydrant wc',\n",
              " 'failure to post calorie information',\n",
              " 'bwso referral to water quality for dep internal use only qbwso',\n",
              " 'snow or ice',\n",
              " 'e street obstruction',\n",
              " 'bikes in buildings',\n",
              " 'going out of business',\n",
              " 'unauthorized film shoot',\n",
              " 'utility cover blocked',\n",
              " 'sidewalk grating  missing',\n",
              " 'exposure unnecessary',\n",
              " 'noise: construction equipment  for dep internal use only ync',\n",
              " 'no dial tone',\n",
              " 'graffiti dirty condition',\n",
              " 'puddle on roof',\n",
              " 'not certified',\n",
              " 'illegal air conditioning during drought cjo',\n",
              " 'plumbing',\n",
              " 'contrary to ll  handicapped access',\n",
              " 'cp employee behaviorbcc',\n",
              " 'conflict monitor',\n",
              " 'monkey',\n",
              " 'broken window',\n",
              " 'a sweeping missed',\n",
              " 'flood light lamp dayburning',\n",
              " 'remove stump',\n",
              " 'e throwout',\n",
              " 'do not block the box',\n",
              " 'fire globe missing',\n",
              " 'bracket arm broken',\n",
              " 'vocational or trade school',\n",
              " 'hitting phone cable lines',\n",
              " 'swimming pool cover',\n",
              " 'professional photographer',\n",
              " 'warning buzzer',\n",
              " 'taste',\n",
              " 'outlet cover',\n",
              " ' dsny spillage',\n",
              " 'ferret',\n",
              " 'fountain  over  gallons',\n",
              " 'no idling',\n",
              " 'traffic camera',\n",
              " 'glassware hanging',\n",
              " 'blocked  atm',\n",
              " 'noise: other noise sources use comments nzz',\n",
              " 'noise: loud music daytime mark date and time nn',\n",
              " 'waterwayfloatables ihc',\n",
              " 'suspected street cut',\n",
              " 'controller cabinet',\n",
              " 'facility unregistered',\n",
              " 'defective hardware  dep',\n",
              " 'broken lock',\n",
              " 'lane control signal',\n",
              " 'fire alarm lamp cycling',\n",
              " 'licensee complaint',\n",
              " 'caps & chains missingfire dept use only wkmc',\n",
              " 'bracket arm loose',\n",
              " 'bracket arm bent',\n",
              " 'i street condition icenon storm',\n",
              " 'beach pool sauna unpermitted',\n",
              " 'time clock maladjusted',\n",
              " 'process server',\n",
              " 'milk not pasteurized',\n",
              " 'asphalt around valve box',\n",
              " 'manhole sunken damaged raised sb',\n",
              " 'detector sensor',\n",
              " 'flooded',\n",
              " 'dead end signal',\n",
              " 'defective catch basin',\n",
              " 'flashing hazard',\n",
              " 'e litter surveillance',\n",
              " 'damaged toilet sink',\n",
              " 'equipment malfunction',\n",
              " 'pedestrian sign',\n",
              " 'glassware missing',\n",
              " 'lane station',\n",
              " 'on structure',\n",
              " 'condulet cover',\n",
              " 'defective manhole cover',\n",
              " 'utility company',\n",
              " 'blocking sidewalk',\n",
              " 'asphalt around manhole cover',\n",
              " 'flood light lamp dim',\n",
              " 'reflector louvre',\n",
              " 'too few on duty',\n",
              " 'plants noise related problems pn',\n",
              " 'ticket scalping',\n",
              " 'manhole cover missing emergency for dep internal use only wf',\n",
              " 'technician unlicensed',\n",
              " 'chained bike',\n",
              " 'box cover']"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 34
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "N5YdojXCtiNf"
      },
      "source": [
        "descriptors_not_in_sample_df = df[df.descriptor.isin(descriptors_not_in_sample)]\n",
        "random_df = pd.concat([random_df, descriptors_not_in_sample_df])\n",
        "random_df.reset_index(inplace=True)"
      ],
      "execution_count": 35,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "yH94tmTEPgUm"
      },
      "source": [
        "random_X = random_df.descriptor\n",
        "random_y = random_df.agency"
      ],
      "execution_count": 36,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "-mOrsN1TQRPr"
      },
      "source": [
        "r_descriptors = random_X\n",
        "tokenizer = Tokenizer() \n",
        "\n",
        "tokenizer.fit_on_texts(r_descriptors) \n",
        "r_descriptor_onehot = tokenizer.texts_to_matrix(r_descriptors, mode='binary')\n",
        "word_index = tokenizer.word_index"
      ],
      "execution_count": 185,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "Q5Ve_W4NQX97"
      },
      "source": [
        "r_agency = random_y\n",
        "\n",
        "le = preprocessing.LabelEncoder() \n",
        "le.fit(r_agency)\n",
        "r_agency_cat = le.transform(r_agency)  \n",
        "r_agency_onehot = to_categorical(r_agency_cat)"
      ],
      "execution_count": 186,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "c9MwlTriRHt1"
      },
      "source": [
        "test_index = random.sample(range(1,10000), 1500)\n",
        "\n",
        "r_test = r_descriptor_onehot[test_index]\n",
        "r_label_test = r_agency_onehot[test_index]"
      ],
      "execution_count": 187,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "md8FyuC0RVLw",
        "outputId": "6475a6fc-d602-4ee2-8393-3cd80b8867b9"
      },
      "source": [
        "random_results = model1.evaluate(r_descriptor_onehot, r_agency_onehot, batch_size=32)"
      ],
      "execution_count": 40,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "3622/3622 [==============================] - 6s 2ms/step - loss: 1.1399 - acc: 0.6294\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "COFR7H471iEo"
      },
      "source": [
        "The model had 62% accuracy on the random sample. This score can be improved by testing different parameters in the model."
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "9TZA4Zl9x2EZ"
      },
      "source": [
        "## Parameter Tuning"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "qpprv5Nh74uC"
      },
      "source": [
        "The function below allows us to create a model and plot its learning history, given the data and a dictionary of parameters."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "t1XAcgaVhHSI"
      },
      "source": [
        "def create_model(X_train, y_train, X_test, y_test, params):\n",
        "    \n",
        "    \"\"\"Takes in training and test data and a dictionary with activation\n",
        "       functions for each of 3 layers, an optimizer, a loss function, \n",
        "       number of epochs and batch size. Plots the model's accuracy and loss\n",
        "       curves for the training and validation data, and prints the accuracy \n",
        "       score for the random sample data. Returns the history and the model.\"\"\"\n",
        "\n",
        "    model = models.Sequential()\n",
        "    model.add(layers.Dense(50, activation=params['activation 1'], \n",
        "                           input_shape=(test.shape[1],)))\n",
        "    model.add(layers.Dense(25, activation=params['activation 2']))\n",
        "    model.add(layers.Dense(14, activation=params['activation 3']))\n",
        "\n",
        "    opt = params['opt']\n",
        "    model.compile(optimizer=opt,\n",
        "              loss=params['loss'],\n",
        "              metrics=['acc'])\n",
        "    \n",
        "    es = keras.callbacks.EarlyStopping(monitor='val_loss', patience=5)\n",
        "    history = model.fit(X_train,\n",
        "                    y_train,\n",
        "                    epochs=params['epochs'],\n",
        "                    batch_size=params['batch_size'],\n",
        "                    validation_split=0.20)\n",
        "    \n",
        "    print(model.summary)\n",
        "\n",
        "    print('Test Accuracy:')\n",
        "    model.evaluate(X_test, y_test, batch_size=params['batch_size'])\n",
        "    \n",
        "    print('Random Sample Accuracy:')\n",
        "    model.evaluate(r_test, r_label_test, batch_size=params['batch_size'])\n",
        "    \n",
        "    plot_loss_acc(history)\n",
        "    \n",
        "    return history, model"
      ],
      "execution_count": 41,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "4jTnHo089K9R"
      },
      "source": [
        "We'll use the function to test out several different sets of parameters, stored in a dictionary as shown below."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "e6QqNu9vI5Bb"
      },
      "source": [
        "model2_params = {'activation 1': 'relu', \n",
        "                 'activation 2': 'softmax', \n",
        "                 'activation 3': 'softmax', \n",
        "                 'opt': keras.optimizers.Adam(lr=0.0001),\n",
        "                 'loss': 'categorical_crossentropy',\n",
        "                 'epochs': 75, \n",
        "                 'batch_size': 16, \n",
        "                 }"
      ],
      "execution_count": 42,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 1000
        },
        "id": "qznZkA3tKeWi",
        "outputId": "386dbdae-bbe8-4ebb-b64b-5b63084b2047"
      },
      "source": [
        "model2_history, model2 = create_model(train, label_train, test, label_test, model2_params)"
      ],
      "execution_count": 43,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Epoch 1/75\n",
            "575/575 [==============================] - 1s 2ms/step - loss: 2.5713 - acc: 0.2788 - val_loss: 2.4859 - val_acc: 0.5113\n",
            "Epoch 2/75\n",
            "575/575 [==============================] - 1s 2ms/step - loss: 2.3708 - acc: 0.5215 - val_loss: 2.2639 - val_acc: 0.5109\n",
            "Epoch 3/75\n",
            "575/575 [==============================] - 1s 2ms/step - loss: 2.1587 - acc: 0.5215 - val_loss: 2.0840 - val_acc: 0.5109\n",
            "Epoch 4/75\n",
            "575/575 [==============================] - 1s 2ms/step - loss: 2.0064 - acc: 0.5512 - val_loss: 1.9529 - val_acc: 0.5527\n",
            "Epoch 5/75\n",
            "575/575 [==============================] - 1s 2ms/step - loss: 1.8843 - acc: 0.5705 - val_loss: 1.8416 - val_acc: 0.5748\n",
            "Epoch 6/75\n",
            "575/575 [==============================] - 1s 2ms/step - loss: 1.7787 - acc: 0.6084 - val_loss: 1.7449 - val_acc: 0.6449\n",
            "Epoch 7/75\n",
            "575/575 [==============================] - 1s 2ms/step - loss: 1.6856 - acc: 0.6911 - val_loss: 1.6580 - val_acc: 0.6989\n",
            "Epoch 8/75\n",
            "575/575 [==============================] - 1s 2ms/step - loss: 1.6003 - acc: 0.7154 - val_loss: 1.5761 - val_acc: 0.7141\n",
            "Epoch 9/75\n",
            "575/575 [==============================] - 1s 2ms/step - loss: 1.5185 - acc: 0.7258 - val_loss: 1.4970 - val_acc: 0.7158\n",
            "Epoch 10/75\n",
            "575/575 [==============================] - 1s 2ms/step - loss: 1.4416 - acc: 0.7267 - val_loss: 1.4252 - val_acc: 0.7167\n",
            "Epoch 11/75\n",
            "575/575 [==============================] - 1s 2ms/step - loss: 1.3729 - acc: 0.7271 - val_loss: 1.3610 - val_acc: 0.7171\n",
            "Epoch 12/75\n",
            "575/575 [==============================] - 1s 2ms/step - loss: 1.3115 - acc: 0.7278 - val_loss: 1.3034 - val_acc: 0.7180\n",
            "Epoch 13/75\n",
            "575/575 [==============================] - 1s 2ms/step - loss: 1.2563 - acc: 0.7280 - val_loss: 1.2515 - val_acc: 0.7180\n",
            "Epoch 14/75\n",
            "575/575 [==============================] - 1s 2ms/step - loss: 1.2065 - acc: 0.7286 - val_loss: 1.2044 - val_acc: 0.7202\n",
            "Epoch 15/75\n",
            "575/575 [==============================] - 1s 2ms/step - loss: 1.1613 - acc: 0.7294 - val_loss: 1.1618 - val_acc: 0.7202\n",
            "Epoch 16/75\n",
            "575/575 [==============================] - 1s 2ms/step - loss: 1.1200 - acc: 0.7619 - val_loss: 1.1220 - val_acc: 0.7681\n",
            "Epoch 17/75\n",
            "575/575 [==============================] - 1s 2ms/step - loss: 1.0780 - acc: 0.7784 - val_loss: 1.0752 - val_acc: 0.7711\n",
            "Epoch 18/75\n",
            "575/575 [==============================] - 1s 2ms/step - loss: 1.0160 - acc: 0.7817 - val_loss: 1.0077 - val_acc: 0.7737\n",
            "Epoch 19/75\n",
            "575/575 [==============================] - 1s 2ms/step - loss: 0.9566 - acc: 0.7845 - val_loss: 0.9574 - val_acc: 0.7746\n",
            "Epoch 20/75\n",
            "575/575 [==============================] - 1s 2ms/step - loss: 0.9099 - acc: 0.7852 - val_loss: 0.9146 - val_acc: 0.7759\n",
            "Epoch 21/75\n",
            "575/575 [==============================] - 1s 2ms/step - loss: 0.8692 - acc: 0.7862 - val_loss: 0.8767 - val_acc: 0.7781\n",
            "Epoch 22/75\n",
            "575/575 [==============================] - 1s 2ms/step - loss: 0.8327 - acc: 0.7884 - val_loss: 0.8420 - val_acc: 0.7781\n",
            "Epoch 23/75\n",
            "575/575 [==============================] - 1s 2ms/step - loss: 0.7994 - acc: 0.7885 - val_loss: 0.8102 - val_acc: 0.7781\n",
            "Epoch 24/75\n",
            "575/575 [==============================] - 1s 2ms/step - loss: 0.7688 - acc: 0.7893 - val_loss: 0.7807 - val_acc: 0.7789\n",
            "Epoch 25/75\n",
            "575/575 [==============================] - 1s 2ms/step - loss: 0.7404 - acc: 0.7904 - val_loss: 0.7533 - val_acc: 0.7811\n",
            "Epoch 26/75\n",
            "575/575 [==============================] - 1s 2ms/step - loss: 0.7141 - acc: 0.7906 - val_loss: 0.7283 - val_acc: 0.7811\n",
            "Epoch 27/75\n",
            "575/575 [==============================] - 1s 2ms/step - loss: 0.6898 - acc: 0.7907 - val_loss: 0.7046 - val_acc: 0.7815\n",
            "Epoch 28/75\n",
            "575/575 [==============================] - 1s 2ms/step - loss: 0.6670 - acc: 0.8221 - val_loss: 0.6825 - val_acc: 0.8316\n",
            "Epoch 29/75\n",
            "575/575 [==============================] - 1s 2ms/step - loss: 0.6458 - acc: 0.8464 - val_loss: 0.6626 - val_acc: 0.8342\n",
            "Epoch 30/75\n",
            "575/575 [==============================] - 1s 2ms/step - loss: 0.6260 - acc: 0.8473 - val_loss: 0.6430 - val_acc: 0.8355\n",
            "Epoch 31/75\n",
            "575/575 [==============================] - 1s 2ms/step - loss: 0.6073 - acc: 0.8483 - val_loss: 0.6252 - val_acc: 0.8355\n",
            "Epoch 32/75\n",
            "575/575 [==============================] - 1s 2ms/step - loss: 0.5898 - acc: 0.8485 - val_loss: 0.6082 - val_acc: 0.8359\n",
            "Epoch 33/75\n",
            "575/575 [==============================] - 1s 2ms/step - loss: 0.5733 - acc: 0.8488 - val_loss: 0.5922 - val_acc: 0.8359\n",
            "Epoch 34/75\n",
            "575/575 [==============================] - 1s 2ms/step - loss: 0.5579 - acc: 0.8488 - val_loss: 0.5771 - val_acc: 0.8368\n",
            "Epoch 35/75\n",
            "575/575 [==============================] - 1s 2ms/step - loss: 0.5433 - acc: 0.8492 - val_loss: 0.5635 - val_acc: 0.8372\n",
            "Epoch 36/75\n",
            "575/575 [==============================] - 1s 2ms/step - loss: 0.5298 - acc: 0.8493 - val_loss: 0.5500 - val_acc: 0.8372\n",
            "Epoch 37/75\n",
            "575/575 [==============================] - 1s 2ms/step - loss: 0.5168 - acc: 0.8493 - val_loss: 0.5378 - val_acc: 0.8372\n",
            "Epoch 38/75\n",
            "575/575 [==============================] - 1s 2ms/step - loss: 0.5047 - acc: 0.8493 - val_loss: 0.5254 - val_acc: 0.8372\n",
            "Epoch 39/75\n",
            "575/575 [==============================] - 1s 2ms/step - loss: 0.4932 - acc: 0.8493 - val_loss: 0.5142 - val_acc: 0.8372\n",
            "Epoch 40/75\n",
            "575/575 [==============================] - 1s 2ms/step - loss: 0.4824 - acc: 0.8493 - val_loss: 0.5037 - val_acc: 0.8372\n",
            "Epoch 41/75\n",
            "575/575 [==============================] - 1s 2ms/step - loss: 0.4718 - acc: 0.8493 - val_loss: 0.4936 - val_acc: 0.8372\n",
            "Epoch 42/75\n",
            "575/575 [==============================] - 1s 2ms/step - loss: 0.4617 - acc: 0.8493 - val_loss: 0.4839 - val_acc: 0.8372\n",
            "Epoch 43/75\n",
            "575/575 [==============================] - 1s 2ms/step - loss: 0.4520 - acc: 0.8493 - val_loss: 0.4742 - val_acc: 0.8372\n",
            "Epoch 44/75\n",
            "575/575 [==============================] - 1s 2ms/step - loss: 0.4424 - acc: 0.8494 - val_loss: 0.4650 - val_acc: 0.8372\n",
            "Epoch 45/75\n",
            "575/575 [==============================] - 1s 2ms/step - loss: 0.4333 - acc: 0.8494 - val_loss: 0.4562 - val_acc: 0.8377\n",
            "Epoch 46/75\n",
            "575/575 [==============================] - 1s 2ms/step - loss: 0.4247 - acc: 0.8495 - val_loss: 0.4480 - val_acc: 0.8377\n",
            "Epoch 47/75\n",
            "575/575 [==============================] - 1s 2ms/step - loss: 0.4167 - acc: 0.8495 - val_loss: 0.4401 - val_acc: 0.8377\n",
            "Epoch 48/75\n",
            "575/575 [==============================] - 1s 2ms/step - loss: 0.4091 - acc: 0.8495 - val_loss: 0.4328 - val_acc: 0.8377\n",
            "Epoch 49/75\n",
            "575/575 [==============================] - 1s 2ms/step - loss: 0.4020 - acc: 0.8495 - val_loss: 0.4264 - val_acc: 0.8377\n",
            "Epoch 50/75\n",
            "575/575 [==============================] - 1s 2ms/step - loss: 0.3953 - acc: 0.8495 - val_loss: 0.4194 - val_acc: 0.8377\n",
            "Epoch 51/75\n",
            "575/575 [==============================] - 1s 2ms/step - loss: 0.3890 - acc: 0.8496 - val_loss: 0.4136 - val_acc: 0.8377\n",
            "Epoch 52/75\n",
            "575/575 [==============================] - 1s 2ms/step - loss: 0.3830 - acc: 0.8496 - val_loss: 0.4081 - val_acc: 0.8377\n",
            "Epoch 53/75\n",
            "575/575 [==============================] - 1s 2ms/step - loss: 0.3774 - acc: 0.8496 - val_loss: 0.4025 - val_acc: 0.8377\n",
            "Epoch 54/75\n",
            "575/575 [==============================] - 1s 2ms/step - loss: 0.3718 - acc: 0.8496 - val_loss: 0.3972 - val_acc: 0.8377\n",
            "Epoch 55/75\n",
            "575/575 [==============================] - 1s 2ms/step - loss: 0.3662 - acc: 0.8496 - val_loss: 0.3914 - val_acc: 0.8377\n",
            "Epoch 56/75\n",
            "575/575 [==============================] - 1s 2ms/step - loss: 0.3605 - acc: 0.8537 - val_loss: 0.3856 - val_acc: 0.8473\n",
            "Epoch 57/75\n",
            "575/575 [==============================] - 1s 2ms/step - loss: 0.3552 - acc: 0.8629 - val_loss: 0.3805 - val_acc: 0.8538\n",
            "Epoch 58/75\n",
            "575/575 [==============================] - 1s 2ms/step - loss: 0.3500 - acc: 0.8652 - val_loss: 0.3758 - val_acc: 0.8560\n",
            "Epoch 59/75\n",
            "575/575 [==============================] - 1s 2ms/step - loss: 0.3453 - acc: 0.8692 - val_loss: 0.3712 - val_acc: 0.8612\n",
            "Epoch 60/75\n",
            "575/575 [==============================] - 1s 2ms/step - loss: 0.3407 - acc: 0.8730 - val_loss: 0.3670 - val_acc: 0.8642\n",
            "Epoch 61/75\n",
            "575/575 [==============================] - 1s 2ms/step - loss: 0.3365 - acc: 0.8760 - val_loss: 0.3628 - val_acc: 0.8664\n",
            "Epoch 62/75\n",
            "575/575 [==============================] - 1s 2ms/step - loss: 0.3323 - acc: 0.8771 - val_loss: 0.3591 - val_acc: 0.8668\n",
            "Epoch 63/75\n",
            "575/575 [==============================] - 1s 2ms/step - loss: 0.3284 - acc: 0.8776 - val_loss: 0.3552 - val_acc: 0.8668\n",
            "Epoch 64/75\n",
            "575/575 [==============================] - 1s 2ms/step - loss: 0.3246 - acc: 0.8776 - val_loss: 0.3519 - val_acc: 0.8668\n",
            "Epoch 65/75\n",
            "575/575 [==============================] - 1s 2ms/step - loss: 0.3211 - acc: 0.8778 - val_loss: 0.3478 - val_acc: 0.8668\n",
            "Epoch 66/75\n",
            "575/575 [==============================] - 1s 2ms/step - loss: 0.3175 - acc: 0.8781 - val_loss: 0.3453 - val_acc: 0.8668\n",
            "Epoch 67/75\n",
            "575/575 [==============================] - 1s 2ms/step - loss: 0.3143 - acc: 0.8789 - val_loss: 0.3418 - val_acc: 0.8668\n",
            "Epoch 68/75\n",
            "575/575 [==============================] - 1s 2ms/step - loss: 0.3111 - acc: 0.8794 - val_loss: 0.3385 - val_acc: 0.8668\n",
            "Epoch 69/75\n",
            "575/575 [==============================] - 1s 2ms/step - loss: 0.3079 - acc: 0.8795 - val_loss: 0.3354 - val_acc: 0.8673\n",
            "Epoch 70/75\n",
            "575/575 [==============================] - 1s 2ms/step - loss: 0.3048 - acc: 0.8795 - val_loss: 0.3335 - val_acc: 0.8673\n",
            "Epoch 71/75\n",
            "575/575 [==============================] - 1s 2ms/step - loss: 0.3021 - acc: 0.8795 - val_loss: 0.3298 - val_acc: 0.8673\n",
            "Epoch 72/75\n",
            "575/575 [==============================] - 1s 2ms/step - loss: 0.2994 - acc: 0.8798 - val_loss: 0.3268 - val_acc: 0.8673\n",
            "Epoch 73/75\n",
            "575/575 [==============================] - 1s 2ms/step - loss: 0.2967 - acc: 0.8801 - val_loss: 0.3242 - val_acc: 0.8673\n",
            "Epoch 74/75\n",
            "575/575 [==============================] - 1s 2ms/step - loss: 0.2940 - acc: 0.8803 - val_loss: 0.3223 - val_acc: 0.8673\n",
            "Epoch 75/75\n",
            "575/575 [==============================] - 1s 2ms/step - loss: 0.2916 - acc: 0.8804 - val_loss: 0.3200 - val_acc: 0.8673\n",
            "<bound method Model.summary of <tensorflow.python.keras.engine.sequential.Sequential object at 0x7fd8b00a1dd8>>\n",
            "Test Accuracy:\n",
            "6461/6461 [==============================] - 10s 2ms/step - loss: 0.3085 - acc: 0.8747\n",
            "Random Sample Accuracy:\n",
            "94/94 [==============================] - 0s 1ms/step - loss: 1.4984 - acc: 0.6320\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAA3sAAAGHCAYAAAAJAkuoAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADh0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uMy4yLjIsIGh0dHA6Ly9tYXRwbG90bGliLm9yZy+WH4yJAAAgAElEQVR4nOzde3ycZZ3//9c1k8m5TdPzISk9n4GeKGcoIgoo6CgKo+iCB1xPKLPuLu7uV1kPq+7qrKIgP3QVURllhVlwBVFWzi1IKdBDWno+pGnTJG3OmSQzc/3+uCfTSZom0zSZySTv5+ORx9yH677vz0zT3PdnrpOx1iIiIiIiIiIjiyvTAYiIiIiIiMjgU7InIiIiIiIyAinZExERERERGYGU7ImIiIiIiIxASvZERERERERGICV7IiIiIiIiI1BOpgMQGSifN+AD7gIWxDe1ADOCIX9Dj3L3AJ+Jrx4DXg+G/G8f4DXHAOuBpfFNtwZD/gdSPPZZ4CLAA/xrMOS/K4VjDLAbmA3MD4b8u04/ahERkdM3Uu+zPm/gH4G/BWbFNx0B/jcY8n9yIDGLDGeq2ZOsFQz5g8A7gc74piLgo8llfN5AMXBj8mEDvQHFr9kUDPmXDfDYtUDVaR62FifRA7hlINcVEREZiJF6nw2G/N8Bbk3a9D0lejJSKdmTkaAK2BBf/tse+z4M/CW94QyqW4FofPmjPm9A/2dFRCTdRvJ9VmREUzNOGSnuA34KLPF5A5cHQ/7n4ts/Bfw98IHeDvJ5A+cCXwfOB9pxmqj8BPh+MOSPxcvMBX4EXA5sBv73FOcqA74DXAG0Ag3AvwdD/t8O5A3Fm7JcBXwX+EegHHg78Kce5RYB38KpBTwGNAJvAF8Phvx74k1Bv4jzWUwH3gLagCDwX8BGkprLAL8EKjjRbOeKYMj/rM8beCL+GRTGj20ALo0fe0X89VNADJgBHAceBP6t67OMx/sR4O+AecAOnM/9CaAu/j7GAh3AT4Mh/2fjzW3ujJ/vymDIv3cgn6eIiJyREXefTUWK8X8O+DRQAtTED306GPL/fSr7RYaSaglkpOhKPsD5g4rPG7gQqMXp83YSnzewBHgJuA64HjgLqAa+B3w/XsYFPAZcDTwQDPnP58Qf6uRzFQPPAR8C7gAW4yQtv/F5AzcM8D19EPgDcD9g49tu6XHdGcCLwHuBQDDkn4uT9F0InBMv9m9AAKf5zcJgyH8e8BDwqWDI39GzuUww5I/iNNuhx/ZrgVfjq+8HfhM/9g/xbdcAPwiG/Mtx+kE04dwgP58U76dwEsB5wHnBkH8l8K+APxjy3wt8JV7UHY+7q7nNfmCFEj0RkYwZiffZPqUY/zuBHwK5wJxgyL8C5wvaD6SyX2SoKdmTESEY8rcCv4qvvs/nDUzBuRn9f30c9mWcBOhoMOR/JRjyW5waJoDP+LyBaTg1a121Xr+Pvz7ay7k+BMyJL68LhvydODVm4HzjORC3Ar8Ihvx7cG42AF6fNzAuqcxngAnx5XsA4h3n/x2ojt8c74jvDwVD/sPx5Z/jJIkDtSnpW91bgb/iNO15IB5DCydqIN+XdFxXMveXYMi/LV72j8Dv4tsfBMI4yd4nAHzewEXAtp4DAoiISPqM0Ptsf1KJv+uL1enA9T5voDgY8v8J8Me397dfZEgp2ZOR5L74qwfnW7NLcL4tPJVV8ddjSdvq4q9uYDmwKGnfsR5lki1PWn7C5w28AazB+QawpN/Ie/B5A/OBKcGQ/4X4pgfjr/nATUlFV8Rf24Mhf+J9BEP+nwVD/vXAEiAvvrkqaX97MOT/3OnGlSTxLW4w5K+JPwSMA37p8wa2+LyBCk50fi+Pv6fJODe7brHEz/HJ+Otx4OH45k/4vAF3/DwPnEGsIiIyOEbMfTZFqcT/Ak73hULgv4FjPm/g98CeeLn+9osMKSV7MmIEQ/4twLr46h3AQ8GQP5Kmy7clLV8bDPmXB0P+2cGQf2ow5F90yqNO7VZgos8b2OfzBvbh1IjFkvb15E7hnKfTR9f0sz+cvOLzBgqB/wN8OLWRSzjxUNDb35m+Yun6lrgMp5nLxcCf+wtYRESG1gi7zw6KYMj/Mk73iUdw+sx7gHcDz/q8gXH97c9EzDK6KNmTkaYrwYjhdKDuS9fIYuOTtnU1iYwCrwPbk/aN71Em2dak5fKuBZ83cL3PG7i7nzi6ifdf+BCwLBjyz4r/lOP0swNY4/MGFseXX4+/5sSb1HSd42M+b+BKnIFW2uObZyTtn+zzBgJJl63vOk/8tbf32JclwOT48uPx124JaDDkP8qJGr3kWHJ83sC9Pm+gKF5uHbAlvvvHwGPJA7yIiEhGZf19tj8+b+CffN7AClKI3+cNXA6YYMh/AzAR+FJ8fwkwr7/9gxm3SG+U7MlI8984I0/+TTDk399P2W/jjKo12ecNnB8ftfLq+L57giH/EZwapa4bzHXx1/f2cq6HgK7BQz4Fic7kXwaeP833cBWwNRjyH+qx/RdJy121e/dyoknJJ+PXnQZ8FagMhvzNwH/G978n3pQS4HNAadL51sdfF8Zfe3uPfaniRM3jonjzy8t6Kfe1+OvlPm+ga7TPjwKL4/38unTV7o2j+/sWEZHMGgn32f4swEnGUol/BfAdnzcwJt6PcHN8fyPO6Nf97RcZUsZa238pkWHI5w34gLtwJh3fAbw3GPLvOkWZrsTiGPB614Sv8SGVv4EzpHIY54/6T4H/7GdI6K6k5RDOSJIPx8v9O/A2nD/ih3CmD/hZ/DzPAhfhNOGoAb4TDPm/1yPedwM/wxl989fBkN+ftK8CZ/QxcIacfiQY8n/U5w0sxBm58or4+6sGvhEM+Z+MH2eA23FujjOAXTid2u+IJ4NdfQQfBM7FaaLze+IjjQH7gJtxBoN5L06/g3pgZzDkX5MU3xdwbrpRnM71hcDHcCbjfSUY8l8aL3czTvOfBTj/bruAL8Rvml3nKsFJIDcHQ/4LEBGRtBuJ99l4ubuAz+LUtIEzymhzUpFS4F3xaYf6jN/nDbwd+H84tY1tOLWAW4C7giH/S/3t7/2TFxk8SvZEZFjyeQO/AZ4Nhvz39VtYRERERE6iSdVFZNjweQM/xBmN83Wc4ar/NrMRiYiIiGQvJXsiMpwcwZm4txn4ZjDkr++nvIiIiIicgppxioiIiIiIjEAajVNERERERGQEUrInIiIiIiIyAmV1n72JEyfaWbNmDfj4mHWmBXOZ7Mt5FXtmZGvs2Ro3KPZMGY6xv/baa7XW2kmZjiNb6B6ZfbFna9yg2DMlW2PP1rhheMbe1/0xbcmeMeZq4AeAG/iptfbbPfafhTO/2CScOVputtZW9nXOWbNmsWHDhgHH1NzhTKlSnFs84HNkimLPjGyNPVvjBsWeKcMxdmNMfxM4SxLdI7Mv9myNGxR7pmRr7NkaNwzP2Pu6P6YlJTXGuIF7gGuAJYDPGLOkR7HvAg9aa8/BmUjzW+mITUREREREZCRKV/3jGmCXtXaPtbYD+A3wnh5llgB/iS8/08t+ERERERERSVG6mnHOAA4mrVcC5/co8ybwPpymnl5gjDFmgrW27lQnjdlYoip1IM7k2ExT7JmRrbFna9yg2DMlm2MXERERx3AaoOVLwI+MMbcAzwOHgGjPQsaY24DbAMpnlp90kkgkQu3hWjraO7D0PYdg1xyDdeaU+eSwlY7YDYbcvFwmTptITs5w+lURERERkeGus7OTyspKwuFwt+3DcZCTVGUy9vz8fMrKyvB4PCkfk64n+ENAcmZWFt+WYK2twqnZwxhTDLzfWlvf80TW2vuB+wFWr15te3aO3HtoL+NKxjFhwgSMMX0GFY05uaTb5T7Nt5N56YjdWktdXR3Hq48ze/bsQT//cOrYerqyNfZsjRsUe6Zkc+wiIpJZlZWVjBkzhlmzZnV7Ltcz+Onrei6vrKw8refydKWkrwLzjTGzjTG5wE3A48kFjDETjUmkyF/GGZnztIXD4ZQSPemfMYYJEyac9G2MiIiIiEh/9Fw+eAb6XJ6WZM9aGwE+BzwFbAMettZuNcZ8zRhzfbzYWuAtY8wOYArwzYFeT79Qg0efpYiIiIgMlJ4lB89APsu0NTa11j5hrV1grZ1rrf1mfNtXrLWPx5d/Z62dHy/zCWtte7piG0z19fXce++9p33ctddeS339Sa1WRURERERkAPRcnsZkb7Q41S9VJBLp87gnnniCcePGDVVYIiIiIiKjip7Lh9donCPCnXfeye7du1m+fDkej4f8/HxKS0vZvn07O3bs4L3vfS8HDx4kHA7zhS98gdtuuw2AWbNmsWHDBpqbm7nmmmu45JJLWLduHTNmzOCxxx6joKAgw+9MRERERCR7DMVz+aOhR7PquXxEJ3tv/ebx/gsN0MKbru91+7e//W22bNnCG2+8wbPPPsu73vUutmzZkhg152c/+xnjx4+nra2N8847j/e///1MmDCh2zl27txJMBjkJz/5CR/84Ad55JFHuPnmm4fsvYiIDDZrLVibeAWDKyf7Rl0bzWwsRqS5ldi4fFyafkdEztBIeS5/9JFH+fDNHx6y9zLY9Nd7iK1Zs6bb8Kh33303oVAIgIMHD7Jz586Tfqlmz57N8uXLAVi1ahX79u1LW7wikl1sNEakvZ1oewfR9g5inZ3EOiNEOzuJdXQSi0S6J14xoJ85SAE6o53YWIwG68JGo8SiUWw0ho05P8RsYvnEj7MNe/L5C6dMovyKCwf/A5AhcfjljTQeOAQxS+7aCymaOinTIYmInLFBeS7fvy9t8Q4GJXtDrKioKLH87LPP8vTTT7N+/XoKCwtZu3Ztr8On5uXlJZbdbjdtbW1piVVEBp+1lkhbmM7WNiKtbUTawsQ6I8QiUScRi0SJxefs6SZmsTaeQHUlWclJV9Q5PtbZd7+DYaOXBFCGL+NyQcz5N2tvaFSyJyIjwmA8l/fX32+4GdHJ3qmqdLsMxaSIY8aMoampqdd9DQ0NlJaWUlhYyPbt23n55ZcH7boiMjxYawnXHafp4GGaqg4TaW4bvYmOMWCcoaKNW+OBZZO8cWMTy+31jRmMRERGiuTn8nRNTK7n8hGe7GXChAkTuPjii1m2bBkFBQVMmTIlse/qq6/mvvvuY/HixSxcuJALLrggg5GKyGCx1tJWU0fTgSqaKg8TDadx5hhjcOd6cOfn4c7NxZ3rweXx4MrNwZ3jweXJAZdx5uYxJ1770x4JY9wuCvKKMG43xu3CuFy43G7nfC4XxrgS203StsT1JGvllSjZE5Hsp+dyJXtD4qGHHup1e15eHk8++WSv+7r65U2cOJEtW7Yktn/pS18a9PhEZHBE2ttp3HuQ+t376WxqOWU5V64HT2EBOYUF5BTm4/Z4cOXk4MrJweS4nSZzPXIjY5ISqK4ftwvjduOKJ1+unBxcuZ4hSayaO5oBKM4tHvRzy/CXXLPX0dCEjcWc31MRkSwz2M/l0d66XgxjSvZERE6DtZa22mPU79xHc+VhZ0CSHtx5uRTPmErO1FJyJ5YwtnBkzNUjo4c7LxdXfh6xcDs2FqOjqYW8kjGZDktERE6Tkj0RkRTEIhEa9x+ifufeXpu1uTw5jD2rjDEzp1MwcQLGZdh16CB1uw9hO4/Q3NpGc2uYtnB7WrvwxWIxWtvaaWkN09wapqW1jfaOzn6P64w6ZTxuz6DEMX/WDD7z0XcPyrkkPTwlxbTHmyS3NzQq2RMRyUJK9kRETiEWidBWe5yWw0dp2HuAWC9JUv74cYybN4uC6VPYsusg655cx7rXtvHyxu1UHqnNQNTD09suOlfJXi+MMeXAg8AUnDkx7rfW/qBHmbXAY8De+KZHrbVfG+rYPCVFtFfXAfF+ezNnDPUlRURkkCnZExFJ0lZ3nKaDVbTVHCN8rL7XkTSN283Ys8rIK5/Oi1v3EPrxI/zhL3/leENzBiKWLBcB/s5au9EYMwZ4zRjzZ2ttRY9yL1hr05ote0pO9NfUIC0iItlJyZ6ISNyxt3ZT8/rWU+73FBUybv4s3jjawE9/9UeefHYDLa0nz8nTpaggj/IZkygpLqKoMJ+iwnwK8vNwudI3UqXBUFSYR1FBPkVF+RQXFpCX6+l3QM72SAcAeTm5gxJH2TTN09Yba+1h4HB8uckYsw2YAfRM9tLOM1bJnohItlOyJyIC1FXspHbTtpO255aMoXDSBIqmT2FfYxtf/I8H+ONzr/V6julTxnPx6qVctGoxF61czOw5k8jJcWfliJYajTP9jDGzgBXAK73svtAY8yZQBXzJWnvqbyXiYjaW+HcciHBejJaOTnJdLmhto7Gl3pnKIwucyfvOpGyNGxR7pgz32GM21uvolTEbH9zs5DHOhr1Mx366f9s1jnKGFRc7D1JVVVXccMMNvZZZu3YtGzZs6PM83//+92ltbU2sX3vttdTX1w9eoCIjlLWW2s3buyV6+RNKmX7JeczzXs3sa64gWj6Dv/vhb1l9/RdOSvTmzZrO33/qBtY9+j32vvgAv/7BP/DZj17HimXzyMkZ2sliZeQwxhQDjwBftNb2rEbbCJxlrT0X+CHwP32c5zZjzAZjzIba2oH3Gf3yt37BBdfcyQVfe4DX9h8BoFPNlEVkhCsZWwI4z+Uf/MAHey3z9re9ndc29P6lb5cf/OAH3Z7L3/2ud2fsuTw7vqIbBaZPn87vfve7AR///e9/n5tvvpnCwkIAnnjiicEKTWTEstZS82YFx7fvTmwrnDyRouVLeWnzTtY9/DQvb9zO+o3buo1gaYzhI+97G1/8uJel82dqAnE5I8YYD06i92tr7aM99ycnf9baJ4wx9xpjJlprT8rmrLX3A/cDrF692g60ZjbcFuFw9XEA9tTUc8HcGbhaOimenl01vdlaM52tcYNiz5ThGrvLuHC7evniM14r1uu+DHO73JSXlfPII4/0XsA4zwF9xX73D+7mox/5KGOKnVGMTzWf30C4jOu0/r1VszfI7rzzTu65557E+l133cU3vvENrrzySlauXMnZZ5/NY489dtJx+/btY9myZQC0tbVx0003sXjxYrxeL21tbYlyn/70p1m9ejXnnH0Od911FwB33303VVVVXHHFFVxxxRUAzJo1i65vdQOBAMuWLWPZsmV8//vfT1xv8eLFfPKTn2Tp0qW84x3v6HYdkZEufLyBw+te65boVUbhb+5/nGnnf4TrP/6vfPveh3n25U3dEr13XLqSV3//A376nS+ybMFZSvTkjBjnF+i/gG3W2sApykyNl8MYswbn3l03lHEtmlueWN5T43wbrX57IpJt0vVcvnTpUr761a8Cw++5fETX7OXOu27Izt2x6/e9br/xxhv54he/yGc/+1kAHn74YZ566iluv/12xo4dS21tLRdccAHXX3/9KR8Sf/zjH1NYWMi2bdvYtGkTK1euTOz75je/yfjx4+no7OCqq65i06ZN3H777QQCAZ555hkmTpzY7VyvvfYaP//5z3nllVew1nL++edz+eWXU1pays6dOwkGg/zkJz/hgx/8II888gg333zzIH1CIsOPjVmaq45w/K09tNWceFaOxmL8dut+vvfos3R0Rno9dsXSuXzjSx/lqktX9rpfZIAuBj4CbDbGvBHf9k/ATABr7X3ADcCnjTERoA24ydqhna1x8bykZO+oU8OnZE9EzsRIfi6PRqNceeWVw/K5fEQne5mwYsUKjh49SlVVFTU1NZSWljJ16lTuuOMOnn/+eVwuF4cOHaK6upqpU6f2eo7nn3+e22+/HYBzzjmHc845J7Hv4Ycf5v777ycSiXD48GEqKiq67e/pxRdfxOv1UlRUBMD73vc+XnjhBa6//npmz57N8uXLAVi1ahX79u0bpE9h5LGxGJFwOx1NjdhoDJc7jI3FEj+xSBQbjWKjMWKxKMROfg6z0RixaI9y1mKtdYb3H8JHt0jMSWAaXH3/l0/EgpMYYS22t8As3WIf9OfOpHPHYjEsFpc584YIsY5Oou0d3bZVHmvkq394mb++tT+xze12ce7iOVy4chEXrlrCRSsXUzZtYs/TiZwxa+2LQJ/Vw9baHwE/Sk9Ejl5r9hoasdaqNltEsoaey5XsDYkPfOAD/O53v+PIkSPceOON/PrXv6ampobXXnsNj8fDrFmzCIdPPVz7qezdu5fvfve7vPrqq4wtGcvHbv3YgM7TJS8vL7HsdrtHdTPOSLidtpo6Im3tRMPtRMJhIuF2Im1hIm1houH2TIc46p08ltcZMob/O1jLvzz4BM1J0yesXDaPn/3HHSyZP3OwryiSNc6aMYmC/Fzawh3UNbfR0BqmpDCfzpZWcouLMh2eiEjK0vFcXlpayi233DIsn8tHdLJ3qirdLl1DwQ5259Abb7yRT37yk9TW1vLcc8/x8MMPM3nyZDweD8888wz79+/v8/jLLruMhx56iLe97W1s2bKFTZs2AdDY2EhRURElJSUcqT7CH//4x0Rb4DFjxtDU1HRSdfGll17KLbfcwp133om1llAoxC9/+ctBfb/ZLtIWZu+TzxBL6pclI5cr18O4ubN49M1dfPG+E2NhuN0uvvyZG/nyZz6IJ0uGlxcZKm63m3mzprF5u3O/2lNTz4qzptJe36hkT0QGJPm5fKiewXuTjufy6upqnnzySdauXQsMr+dyPdEMgaVLl9LU1MSMGTOYNm0aH/7wh7nuuus4++yzWb16NYsWLerz+E9/+tPceuutLF68mMWLF7Nq1SoAzj33XFasWMGiRYsoKy/joosuShxz2223cfXVVzN9+nSeeeaZxPaVK1dyyy23sGbNGgA+8YlPsGLFCjXZTFK/e39KiZ47LxdXfi7G7cadk4NxuU78uF243G6M241xuzC9NDk07h7lXC6My4AxTrOoIWwaFY443zTl5+T3W7YrlkRMxpmYu5eCifgTxw2WrusaQ2ukDTAUeQoH4bzgKSzkgUee5vNfuz+xecGcGTzwXT+rz1lw5tcQGSEWzJ2RSPZ2Hz2R7I0pm5bhyEREUjfYz+UrVzl99pKfy8vLy7n44osTxwyn53IzxH28h9Tq1attz/nntm3bxuLFi1M6Pp3fKgy2dMZ+Op9pKobTZM3WWvb8/mkirU5VeXHZNPJKxuDOzyOnIJ+cxGs+xu0aVrGfjmyNGwY/9l/87mlu+/LdiX6Gq86ex5O/+Drjxg7+Z6PPfXAZY16z1q7OdBzZord75Om46+4H+be7/xuAj1y0jH+49kKKy6cx4+LzBivEITMcf39Tka1xg2LPlOEe+6meIfUMPnC9faZ93R9VsyejWmt1TSLRc+V6mHbhSlzu7PvDI6l58NH/65borVg6lyceGJpETyTbLZgzI7Gs6RdERLKTkj0Z1Rr2HEgsl8wqTznRi0ajtLa109waJtxjdEdroaGphaO19VTXHqe6tp6augaaWtpobm2jpbWdltY2wu3p6yMYtfFvoczpJ7KRaJSW1jCtbWGaW8I0t4bpjPQ+PcFwF43GEsvLl8zhyV98ndISJXoivVk4tyyx3JXsdTa1EItEcOXo8UFEJBvor7WMWtH2DporjwBw6HgTb+49woENb7F7/2H2HDjM3oNHaGntPgpnNBalLdxBW7ijt1NKljhn8Wye/MXXGT9uTKZDERm25sycQk6Om0gkyuH6ZlrbOynM89De0ETBhNJMhyciIikYkcme5gEaPNncp7MvzS1t/O8jf+bJP73Eup2V7K9T06TR4pLzlvLwPV9mQunYTIciMqx5PDnMmTmVHXsOAbC3tp6lMybRXt+oZE9EUqbn8sEzkOfyEZfs5efnU1dXx4QJE/SLdYastdTV1ZGf3/8Ijtnkngd/zz9+62d0dJ5ZU8SiwnyKC/PJz8s9aSDNwoJ8pk4qZfLEcUydWMqkCSWMG1tEYUE+xUX5FBcWkJ/nSdvvaGun0y+x0FNw2se6jIuionyKErHnk+vxDHaIpzTYnc9zctQnUyRVC+bOSCR7e47Gk70GfTkmIqnRc/ngGehz+YhL9srKyqisrKSmpqbfsjHr9N9x9TJM/nCXrtjz8/MpKyvrv2CWiEajfOV7vzwp0SvIz+WiVUtYNLeMOTOnMfesacyZOfWkZn4tHS0UFuQxaex4XK7s+b0Z7qN19SUn5iRnStJE0m/hnOn8b3x5d81xQIO0iEjqTvVcrmfwgRnIc/mIS/Y8Hg+zZ89OqWw2PwBnc+yZtGv/YZpanFqugtwcblqzhLdfuoLrbr6e/Lzcfo9v7nD+y2RToiciMlC9DdLSXt+oZlkikpJTPZdn83NstsU+4pI9kb68uW1vYnnlWVPxX30+M99+SUqJnojIaLNgzvTE8t54shfr6CTSFsZTePrNwkVEJL1UPSGjyhtbdyeWF0+bSO7YYvI10ICISK/mzZ6eqME7UNdIR8SZxqWjoSmTYYmISIqU7MmosimpZm/htPGUzJmppkgiIqdQWJDHrLLJAMSsZX9dAwDtjUr2RESygZI9GTWstd1r9qZPYuys8gxGJCIy/C2ed+Lv5O6jziAtHfVK9kREsoGSPRk1Dh89xtFjzrfShbkeFi6cTU5+XoajEhEZ3hbNPZHsdfXbU82eiEh2ULIno0by4CwLp45nTPnUDEYjIpIdFs+bmVhOjMjZ0DigyX1FRCS9lOzJqPH6lp2J5YXTJlA8XcmeiEh/FvUy/YKNRIm0tmUqJBERSZGSPRk1Nr7+VmJ52ZwZ5I4pymA0IiLZYVFSn719tQ1Eos6Ewu0akVNEZNhTsiejxpvbTzTjXLViUQYjERHJHiVjipg+ZTwAHZEoh447SV57Q2MmwxIRkRQo2ZNRoaGxhf3VxwDIcRlWrDk7wxGJiGSP5EFauppyaq49EZHhT8mejAob/ro5sTxn8nhK1V9PRCRlvU2/oGacIiLDX9qSPWPM1caYt4wxu4wxd/ayf6Yx5hljzOvGmE3GmGvTFZuMfBte3ZJYPnteOcalidRFRFLV2/QLHY3NGpFTRGSYS0uyZ4xxA/cA1wBLAJ8xZkmPYv8CPGytXQHcBNybjthkdHhj667E8opz5mcwEhGR7JNcs7e3zpmv1EajdDa3ZiokERFJQU6arrMG2GWt3QNgjPkN8B6gIqmMBYAmNtkAACAASURBVMbGl0uAqv5OGrMxmjuaBxzUmRybaYo9dZHWNir2H0msL1s1f8AxZOvnnq1xg2LPlGyOXQZf8oice47WY63FGENHY5NGNhYRGcbS1YxzBnAwab0yvi3ZXcDNxphK4Ang872dyBhzmzFmgzFmQ21t7VDEKiNM44Fqdsb7mACcs2R2BqMREck+k8aXMKF0DAAt7R0nRuSs14icIiLDWbpq9lLhAx6w1n7PGHMh8EtjzDJrbSy5kLX2fuB+gNWrV9vi3OIzvvBgnCNTFHv/1r25KzEvVPmU8ZRNOvPBWbL1c8/WuEGxZ0o2xy6DxxjDuYvn8Jd1bwJQUVVL2fixtDdqkBYRkeEsXTV7h4DypPWy+LZkHwceBrDWrgfygYlpiU5GrFhnhDe27E6sL18yN4PRiIhkrxXL5iWWK6qcljWafkFEZHhLV83eq8B8Y8xsnCTvJuBDPcocAK4EHjDGLMZJ9mrSFJ8Mc7FolGi4g0hbG5HWMJ2trURaw0TawthYzPmJxl+tBWux1hKLRNhedaK5rwZnEREZmFXJyd6heLLX2IyNxTAuzeQkIjIcpSXZs9ZGjDGfA54C3MDPrLVbjTFfAzZYax8H/g74iTHmDpzBWm6xGtN5RIt2dBKuO05bzTHaGxqJRaNgLViw1mKjUaLtHUTa27GR6ICvs+3wiWTv3MVzBiN0EZFRZ2Vysne4zvliLRajs7mV3LFq7isiMhylrc+etfYJnIFXkrd9JWm5Arg4XfHI0LPWYiNRIuEwna1tRFrb6Gxpo62pkY7jjUQaW9ISw1uH6xLry5co2RMRGYjZ5VMoLSnmeEMzDa1hquqbmVE6hvaGRiV7IiLD1HAaoEWylI3FaKs9TnPVEdpqjxFt7yDW0Um0o9OpqRsMxuDOyyUnPw9PUSE5hQXkFObjKSjA5LgxLlfSjwFjMMaw91ANze2dAIwfN4ayaeoGKiIyEMYYViyd222QFifZa2JMeT8Hi4hIRijZkwGJdUZorqqmueoILYePEuvoPP2TGEPeuLEUTBxPwYRSXLkejDmRqBm3C3deLu68PFyeHGdfEmstNccaiCQ38bTQ2R6hpTVMS2s48VACTq1ez3OIiEjqViybl/i7uvVQDVctna1BWkREhjEle5KyWCRKy+Fqmg5U0VxVjY3234/OuN2483LxFBaQU1iAp6iAaK4Lz5hCSqdMx+VJ/Vewo6OTNyr2sO61bazbWMHLG7dzpOZ4/wfGnbtY8+uJiJyJlUtPjGjcNUhLu5I9EZFhS8me9MvGLLWbt1O/cy+xSKTXMjkF+RTNmELxtCl4iotw53pw5Xpwud2JMoeO1PHcxm28tHELTS1teFyePq/b2tbO0bp6qmvrqa45Tu3xRs5kzJ5L15w94GNFRKTHIC1VtVhr6WhqxkZjGLdG5BQRGW6U7Em/ju/YzbFtO0/anlsyhjHl0ymeMZW8cWMTTSRbWsPsOHCE3fsPs+fAYd6o2MP617ax/9DRQY+tuKiAMUUF3ba5XS6KivIpLsynqCCfosJ8Lr/gbN71tvMG/foiIqPJnJlTGTe2iPrGFhra2hODtHQ0NZM3bmymwxMRkR6U7EmfIuF26rbuSKx7iosYM3M61bh4bsse9m952al9qzlOdW09h48eO62mladrdvkULly5mItWLeHCVYtZOn8mLs3vJCKSFl2DtDyzfhOQNEhLY5OSPRGRYUjJnvSpdvN2Yp0RmsIdvHqolk2tEZ7+j4cGVEtXkJ/LmnMXsnr5XGZMm0h+Tl6f5XM9OUyeOI4pE8YxZVIpk8aX4DmNPn4iIjL4Vp49/0Syd6hWg7SIiAxjenKWUzq45wAP/vYpnt66l1f2VBGJxlI6LifHzayyKcydOZU5M6exYM4Mzl++kHMXz8HjyaG5oxmA4lzNyyQikm26DdJSVQNAe31jpsIREZE+KNmTkzz38ma+fvdDvPjqVmKnGBCluKiAtRecw8plc5k2aTyTJ45j6qRSJk8cx4wpE8jJcfd6nIiIZLfkQVq2HnIGaQkfq8daq+ltRESGGSV70k1HRyc3fu5bHKs/uUnOiqVzeeflq3jHpSs5f/lCNakUERmFeh2kxRg6W1rJLS7KdHgiIpJET+vSTcWug4lEzxhYMXMq112+ips//j7OmjE5w9GJiEimnWqQlraaY0r2RESGGQ1jKN28vmVXYvnKxbP45Wffx53/+DEleiIikrAieb69+OTqbTXHMhWOiIicgpI96ea1zSfm01syfSITly3EnZebwYhERGS46T65ujNIS1ttXabCERGRU1CyJ91sfPPEnHrLZk9n3LxZmQtGRESGpe4jcjqDtHQ0NhNpb89gVCIi0pOSPUmIRKJs3nkgsb5mzTKMJiwXEZEe5p41jZIxTv+8+tZ2Dtc7U+qoKaeIyPCiJ3lJqNh1gPbOCABTS4o4a9Hcfo4QEZHRyBjDymUn7hFbq+L99mqV7ImIDCdK9iTh1Ve3JpaXzJhEweQJGYxGRESGMw3SIiIy/CnZk4RXX6tILJ+78Cxcbk2MLiIivUsepGVz5VEAwsfqiUUimQpJRER6ULInCa9v25NYPm/F4gxGIiIiw90FKxYmlt84WE24MwLWEq6rz2BUIiKSTMmeANDR1s62A0cS6xdcujKD0YiIyHA3c/pkFsyZAUB7Z5SN+517SGuNpmAQERkulOwJAG9u2EK4MwrA5JJiysqnZTgiEREZ7q66ZEVief2uQ4AGaRERGU6U7AkAr7yyKbF87vzyDEYiIiLZ4qqkViDrdlUCTrJnY7FMhSQiIkmU7AnWWjZu2ZVYX7V8YR+lRUREHJetWYbHkwPAjiPHqGlqxUaitNc3ZjgyEREBJXsCtB9voOJgdWJ9zXlnZzAaERHJFsVFBVy08sSAXuuTavdERCTzlOwJjZVH2Hb4RIf65IlyRURE+vL2pH576+L99jRIi4jI8KBkT9j8xnbaOpx5kSaPH8v0KZpMXUQkHYwx5caYZ4wxFcaYrcaYL/RSxhhj7jbG7DLGbDLGDKvhkq+6tPsgLbGYpa3mGNbaDEYlIiKgZG/Ui4Tbeb1id2J91dnzMxiNiMioEwH+zlq7BLgA+KwxZkmPMtcA8+M/twE/Tm+IfVu+ZA4TS8cCcKyljbeO1BENt9PZ3JrhyERERMneKGCtJdYZobO1jfaGJlqP1tF0oIrjO/Zy9PUtbKuqTZRdqWRPRCRtrLWHrbUb48tNwDZgRo9i7wEetI6XgXHGmGEzP47L5eLKS5Yn1hOjcqopp4hIxuVkOgA5tVg0SjTcTiTcTrS9g1gkgo1EiUUihMOtxCIRmiOGaEcHsY5Ooh0d2FgMG405r9Zio055+mhNU5GU7K1Yqv56IiKZYIyZBawAXumxawZwMGm9Mr7tcF/ni9kYzR3NA47ndI699IIl/Pb3zwNOU86PX7ac+kOHcJeNH/D1z8SZvO9Myta4QbFnSrbGnq1xQ/bFrmRvmGmrPUbNGxW0NzYR6+gc8uvFYpbt3QZnmTfk1xQRke6MMcXAI8AXrbUDnrfAGHMbTlNPymemb87UtRedGMV54/4jtHZ0Yqqd+faMS42IREQyRcneMBI+Vk/lsy87NXGDzLjduDw5uD0eXLkecvLzcBfksb+2kZZ2J6mcPGEcM6ZqcBYRkXQyxnhwEr1fW2sf7aXIISA5cyuLbzuJtfZ+4H6A1atX2+Lc4jOOL5VzLCgvZsn8mVTsPEBnNMZr+45w6YJy3A0dFE6ZeMYxDNRgvP9MyNa4QbFnSrbGnq1xQ/bErmRvmOhoaqHyuR6JnjG483LJKcjHnZeLy5ODK8f5iZgorpwc8guKcOfl4s714M7NxbjdGJfBuFzOj9uFKycn8c1qJBKlManTfMWe9YnllcvmYoxJ23sWERntjPNH97+AbdbawCmKPQ58zhjzG+B8oMFa22cTzky46pIVVOw8AMC6nZVcuqCc5qojGU32RERGOyV7w0CkLUzls+uJtncA4PJ4KL/iQvJKS06ZfHW1F+7vW4X6xmZeeX0z6zZuY/1r2/jrm2/R2tbea9nl6q8nIpJuFwMfATYbY96Ib/snYCaAtfY+4AngWmAX0ArcmoE4+3XVpSv5wc8fA04M0tJSddTphSgiIhmhZC/Dop2dVD73Mp0tTm2bcbuYcdkacsaO4UjNcapr6zlaW0/NsQZaWsM0t7bR3BLmeFMj9Y0tHDvWTHVtPdU1xzl6rIHOzoE3AV1z7sLBelsiIpICa+2LQJ9NKqwzYd1n0xPRwF1y3hLycj20d3Syp6aeI/XNTAU6mprJHZMdzZ1EREYaJXsZZKMxDjz7Clvf2sdbh4+xo7qOfeEI2//zvzlSc3zIJqQdW1yI232iw7wxhqsvX8U1a1cNyfVERGTkKyzI55LzlvJ/LzkVlC/urOSG8xbRUnWU3IVK9kREMkHJXgZtWr+R6++8h8MNQzeEa06Om3MWzebClYu4cNUSLlyxiPLpk4bseiIiMnq987JViWTv/7bt44bzFtFcVU3pwjkZjkxEZHRSspch1lq++O0H+kz0JpaOZcqkUqZMHMekCSWMLS6kqCCfosJ8PHkuxo4p5KypU5kycRyTJ45jysRxFOTnpfFdiIiInPCed1zAP3zrvwB4efchmsIdmJo6Yp0RXB49coiIpJv+8mbIL4NP8tL2/QC4jOGatas4Z8lczlk0m3MWz2bWjMl4+rgxpjpAi4iISLrMLp/K8iVzeKNiD5FojOe2H+Ddy+fRUl3DmLJpmQ5PRGTUUbKXAccbmrnzew8m1m95x/ncd88/ZzAiERGRweG9+iLeqNgDwNMVe51kr6payZ6ISAa4+i8ig+2fvvNzahtaAJg8toh//dJHMxyRiIjI4PC+86LE8ks7D9La0UlzVfWQDTomIiKnlrZkzxhztTHmLWPMLmPMnb3s/09jzBvxnx3GmPp0xZZO6zdu478e/lNi/f/dcAWTZ5VlMCIREZHBs2huOYvmlgMQ7ozy0s5KouF22o83ZDgyEZHRJy3JnjHGDdwDXAMsAXzGmCXJZay1d1hrl1trlwM/BB5NR2zp1NkZ4TP/ck9ife2imbz/vVeccuJ0ERGRbOS9+kTt3tNb9wLQXFWdqXBEREatdPXZWwPsstbuATDG/AZ4D1BxivI+4Kv9nTRmY4mBSgbiTI4diB/89HG27nAGZSnw5PDld1+Mp2zCgOJId+yDSbGnX7bGDYo9U7I5dsk87zsv4lv3/BaA5986QEckSktVNROXLcxwZCIio0u6mnHOAA4mrVfGt53EGHMWMBv4yyn232aM2WCM2VBbWzvogQ6V1rZ2vndfKLH+mStXMWfxLNyaKkFEREaYcxfPZnb5FACa2zt5efchwsfqibSFMxyZiMjoMhxH47wJ+J21NtrbTmvt/cD9AKtXr7aDMfVAOqYvqKg4RHOrc5ObPq6YD1+4jAnz557xtbN56gXFnn7ZGjco9kzJ5tglc4wxeN95EYGfOl9yPr11L5ctnEnTwSpKF2iCdRGRdElXzd4hoDxpvSy+rTc3AcEhjyjNtry1L7F8dtlkCooLKZo6KXMBiYiIDKHkUTmf2b6fSDRG04GqDEYkIjL6pCvZexWYb4yZbYzJxUnoHu9ZyBizCCgF1qcprrTZEu+rBzB/Sikls2diXJr5QkRERqbzzl3A9CnjAahvbWfDvsO01R6js6Utw5GJiIweack2rLUR4HPAU8A24GFr7VZjzNeMMdcnFb0J+I0dgZPxbN6+N7E8f+p4xs4p76O0iIhIdnO5XLz3HUmjclbsA6Dp4Kka9oiIyGBLW9WStfYJa+0Ca+1ca+0349u+Yq19PKnMXdbak+bgy3bWWra8daJmb+ncMnKLizIYkYiIyNBLbsr5l4p9xGJWTTlFRNJI7QjToLq2nrr6JgAKcnM4q2xqhiMSEREZehevXsKk8SUA1DS18ureKsLH6ulobslwZCIio4OSvTRIHpxl/uTx5BUXZi4YERGRNMnJcfOBd12aWH/89Z0Aqt0TEUkTJXtpkNyEc/6UUnIKCzIYjYiISPp85H1vSyz/eeteWto7aDqgfnsiIumgZC8NtuzYl1ieP3U8HiV7IiIySqxcNo/F85xBydo6I/x56z7a6xvpaGzOcGQiIiOfkr006F6zN141eyIiMmoYY/jI+65MrD/++g4AGlW7JyIy5JTsDbFoNErFzgOJ9flTVLMnIiKjy4fesxZXfG7ZV/ce5tDxJvXbExFJAyV7Q2z3gSOE2zsAmFhcQGlRvmr2RERkVJk+ZQJXXbI8sf77N3bS0dhEe31jBqMSERn5lOwNsW4jcU4ZjzsvF1eOO3MBiYiIZMDNSU05f//GTqzVnHsiIkNNyd4Q69Zfb6r664mIyOh0/dvPp2RMEQAH6hp540A1jfsrsdZmODIRkZFLyd4Q6zYSp/rriYjIKFWQn8cN116SWH/s9Z10trTSWl2bwahEREY2JXtDTHPsiYiIOJLn3Htq827CnREa9h7o4wgRETkTSvaGUGtbmN37DwPgMoa5k0tVsyciIqPWhSsXM++saQA0t3fyl4p9NB88TLSjI8ORiYiMTEr2htC2XQcTfRFmThhLvidHNXsiIjJqGWO6DdTyPxt3YGMxGvdrzj0RkaGgZG8I9ZxMHSCnMD9T4YiIiGTczd4rMMYAsH73IQ7UNdCwR005RUSGgpK9IdR92oVSADyFhRmKRkREJPNmTp/MNWtXJ9Yf/us22o83ED7ekMGoRERGJiV7Q2jLjh41ewZyCvIyGJGIiEjm/e3N1yaWQxt3OAO1qHZPRGTQKdkbQj0nVM/Jz8e49JGLiMjo9o5LVzK7fAoAjW3t/HHzHhr3VxKLRjMcmYjIyKLMY4jU1DVQXVsPQL7HTdn4MeQUaXAWERERl8vFbR86Ubv38F8riHV00lx5JINRiYiMPEr2hkjyZOpzJ5fidrk07YKIiEjc37z/SvJyPQBsrqxh66EaGvbs7+coERE5HUr2hkjvI3Eq2RMREQGYOL6EG669JLH+21cqaK2upbO5NYNRiYiMLEr2hkjP/nqAavZERESS/O2HTzTlfGLTbhra2qlX7Z6IyKBRsjdEtu06mFhWzZ6IiMjJ1ixfyPIlcwBoj0R5bOMOGnbv10AtIiKDRMneENlz4EQn81kTSwDV7ImIiCQzxnSr3Xv4rxV0trXTdLAqg1GJiIwcSvaGQFNzK0frnJE4c9wuJo91JlJXzZ6IiEh3N153OSVjigDYX9fIy3sOUb9zX2aDEhEZIZTsDYG9ldWJ5bLSMbhdLozbhTsvN4NRiYiIDD9Fhfl85H1vS6z/ev0WwnXHCR+rz2BUIiIjQ0rJns8beHKoAxlJkptwlo0fC0BOQQHGmEyFJCIiMmx9+iPvTtwjn3/rIPtq6zm+c2+GoxIRyX6p1uwt93kDz/i8gf/weQMrhjSiEWDPgcOJ5bLSMYD664mIiJzK/FnTufaK8xLrv16/lab9h4i0t2cwKhGR7JdqsndbMOS/Avhv4G983sBffN7AP/u8gTlDGFvW2ptUs1feVbOnZE9EROSUbr/1+sTyYxt3UN/SRsOeAxmMSEQk+6Wa7L0Qf60AtgJFwNeAnT5vYL3PG/iMzxvIH4oAs1H3Zpyq2RMREenP2gvO4exFswBo64zw6Ibt1O/ch43ZzAYmIpLFUk32XvZ5Aw8CR4D7gPOANuBXwJeBFuCxIYkwC+3prWavSMmeiIjIqRhjuP2W9yTWH3p5K+GmFloOV/dxlIiI9CXVZG8BcDNQCKwHPglMC4b8fxMM+Z8FfglMH5IIs0wkEmV/1dHEelmpk+ypZk9ERKRvN153GZMnjAPgSEMLT1fs5fgODdQiIjJQOSmWawTuBR4Ihvw7etn/b8ArgxZVFjt4uIZIJArApLGFFOQ6H7H67ImIiPQtPy+X2z50Dd/4YRCAX63bwtVnz6W9oYm8kjEZjk5EJPukmux9IRjy/6JrxecN5ARD/kjXejDkv3PQI8tS3frrxWv1QDV7IiIiqbjtQ9fw7/f9Nx2dEd48eJRNB49SsmMPU887N9OhiYhknVSbcS7zeQO7fd7AFfH1C33ewGafN3DOUAWWrfYeTOqvF592weXJweVJNa8WEREZvaZOKuXG6y5LrP9q3RYa91USbe/IYFQiItkp1WTvfOC8YMj/DEAw5H8BeBfwraEKLFvt7mUkTjXhFBERSd3nkwZq+dPWPVTVNVC/e38GIxIRyU6pJnueYMh/LHlDMOQ/AIwb/JCy295uyZ4GZxERETldy5fM4fLzzwYgGrP8av0W6nfuxUZjGY5MRCS7pNq2cIrPG/hP4A84g7WMBa4FJg9VYNlqz4HDiWVNqC4iIjIwd3zCy3OvbAbgd69u51NrVzDpYBVjZ5VlODIRkeyRas3e74AvAE/hTL3wVHz9t0MUV1ay1mpCdRERkUFw9eWrWDyvHIDWjk4e2bCd4zv2YK0mWRcRSVWqyd5XcCZTbwcMEAbuAb42RHFlpWP1TTQ2twJQmOdhQnwiddXsiYiInB6Xy8UdH/cm1n+1fitNR+toqz3Wx1EiIpIspWQvGPKHgyH/Z4AiYApQHAz5bwcmDGVw2Sa5Vq98QgnGGAA8RUr2RERETpfv+rVMmegMD3C0sYUnN+/m+Ft7MhyViEj2SLVmD583kA+cCywGLvV5A5fh9OFLiTHmamPMW8aYXcaYXuflM8Z80BhTYYzZaox5KNVzDxe7k/rrzSgpchYM5JWWZCgiERGR7JWX5+GzH70usf7Ai5tpqqyiM96KRkRE+pZSsufzBi4GDgCvAc8k/aQ0w6kxxo3T7PMaYAngM8Ys6VFmPvBl4GJr7VLgiym+h2Gj2xx78cFZ8kvH4fZ4MhWSiIgMMZ83kOfzBsbGl/N93sB7fN7ABakeb4z5mTHmqDFmyyn2rzXGNBhj3oj/fGWwYs8Gt33oGooK8wHYWX2MdTsrOb5TtXsiIqlItWbvO4AH6MBJ+iqBTuBwXwclWQPsstbusdZ2AL8B3tOjzCeBe6y1xwGstUdTPPewsaeXaRcKJqulq4jICPcI8HQ84XsWeBR4yecNfCnF4x8Aru6nzAvW2uXxn1HVX378uDHc+oGrEusPvLiZ+t37iXZoknURkf6kOvXCOGAa8OdgyH8pgM8bGA98LMXjZwAHk9YrcSZqT7YAwBjzEuAG7rLW/rGvk8ZsjOaO5hRDONmZHNubXfsPJZbL4yNxusYXDfp1YPBjTyfFnn7ZGjco9kzJ5tgzYApwAXAhzpebDcDLwKeA7/Z3sLX2eWPMrCGML+t9/pbrufeXfyAWi/Hy7kNUHKhmws59TFi6INOhiYgMa6kme1XBkD/s8wYKfN7AXcALOMnfbaRwIzuNWOYDa4Ey4HljzNnW2vrkQsaY2+LXpXxm+SBdenDsPVidWC4fPxaMIXeC5p0XERnh2oIhf9TnDbwLsMA/B0P+e33ewCuDeI0LjTFvAlXAl6y1W/s7YLh9IXomJk0p5vp3rOF//vgyAD9/cRNLZ08jd84UjNt9UvnhFPvpyNa4QbFnSrbGnq1xQ/bFnmqyV+zzBt4JvIQzDUPXJDf7Uzz+EJCcmZXFtyWrBF6x1nYCe40xO3CSv1eTC1lr7wfuB1i9erUtzi1OMYRTG4xzhNs7OFx9HACXMUwrKSa/tISxRUOb7A1G7Jmi2NMvW+MGxZ4p2Rx7GpX5vIEfATfh3B8fi283g3T+jcBZ1tpmY8y1wP/g3B9PMpy/ED1Tn//YuxPJ3lOb9/C3lTWM2X+Y4jmaZF1E5FRSTfb8QAT4KlCM0wSzGifxS8WrwHxjzGycJO8m4EM9yvwP4AN+boyZiNOsM2t6YO+rrE5M9DqtpAhPjpvCyRMzHJWIiKTBg5y4Hz4GdPq8gV8AscE4ubW2MWn5CWPMvcaYidba2l7KDssvRAfDpSvP5R2XruRPL2wkZi33P/s635s1nSkLFmBcvQ9BMFxiP13ZGjco9kzJ1tizNW7InthTTfb+DTgQDPlvAT5+uhex1kaMMZ8DnsLpj/cza+1WY8zXgA3W2sfj+95hjKkAosDfW2vrTvdamaLBWURERqdgyH+Xzxv4C07/9ieBEuBxnAHNzpgxZipQba21xpg1OIOrZc39cTD9y+d9/OmFjQA8uWk3t+2rYuLBKsaepdo9EZHepJrsTaL/kcL6ZK19Aniix7avJC1bnBpE/5lcJ1P2JM2xVxbvr1cwaXwGIxIRkXTweQPnciLRWwB8Aqf1S0pz0Rpjgjj91ScaYypxWtF4AKy19wE3AJ82xkSANuAm29WUZJS5YOWik2r3Fs+fyZiZMzBmsFrNioiMHKlOvfASMKfnRp838MvBDSd7dR+cZQz5pSWaX09EZHT4EXAFUAT8Gbgd+Cbww1QOttb6rLXTrLUea22Ztfa/rLX3xRM9rLU/stYutdaea629wFq7bqjeSDb4l8/7EstPbtrNtp0HaD1Sk8GIRESGr1Rr9pYBm33ewHaglhMDtCwfkqiy0O793Wv21F9PRGT0CIb8d/i8AS8wFdgMPAR8OrNRjUy91e4tXTKXommTMx2aiMiwk2rN3kU4fe2WApfjNDdZi9MvQYC9B0/02SsfP1b99URERo+C+Ou1OF+GfiMY8n+HQeqzJyf7f7d3r93bWrGbttpjGYxIRGR4SjXZewu4tcfPx4AdQxRXVonFYt2bcU4oUX89EZHRI8/nDWwB/gZoB/7o8wYKcPrxyRA4f8Ui3nnZSoBE7V7t5rcyHJWIyPCTajPOTwVD/ud7bvR5A5sGOZ6sdPjoccLtHQCUFOQxadok9dcTERk97gJ+jdMC5uvANJwpGA5mMKYR718+7+Op50+MzLlly04mLJlP4RR1oxAR6ZJqsrfP5w3M7GX7j3CaeI5qPZtwqr+eiMjoEQz5H/F5A78HCoIhf4PPG3ADlwDhDIc2onXV7j31M5yRswAAIABJREFUvNN37+4/v8qSRbOZ+fZLMh2aiMiwkWozzr2n+Dl/iOLKKjv2VCaW1V9PRGRUugr4tc8b2Az8D3BeMORvyXBMI95dd9ycWP6/in28sKGClqrqPo4QERldUk32zCl+BPhV6C+J5QXTxqu/nojIKOLzBm4Cfo8zQMtS4F3AH3zewA0ZDWwUWHX2fHzXr02sf/fJlzn65jZG6TSEIiInSbUZ538EQ/5/7FrxeQP5wGrgvCGJKou8uW0vL26oACDHZfjA2tXqryciMrr8I/C/wF+BVqDw/2/vvuPbqu7/j7+Oh7xjx3bsOLETZy8CSQgkIQQSNoQGXKDF0B+FUkaBQisopYtS2lIK1ECBUihQCgWxvrikQMto2aPsHQLZcRLHK95DlnV+f0hWZJNhiK1rWe/n46GHdK+urt5y9Mjxx+fccwiMfPkp8IiDuWLCry/5f5Q/9QrtHZ18srmGR198h9OmjCG1KN/paCIijutTsRde6AV1AGuAG4Hr+ztUNPnTPf8MPT5sxjjGTRnnYBoREXFAq6fcvaz3ztKSsphe/DxSxozK48IzjueaPz8MwI1Pv8mR+02jePQIh5OJiDivT8VeaUlZ106equrHLFGndlsjnseeD22fumAvsiaOdS6QiIg4IbO0pOxi4C229+ztD2Q4miqGXHrOidz98NNU1TawtbGFu595k0v3mkjGlGFORxMRcdSeXLNXD1w8QLmiwl8ffoZ2bycA00blctDBc0lMS3U4lYiIRFg5cC3wX+D14P3VaAhnxAzLSOWXPzg1tH3ni++z5o2P8Pt29rdqEZHY0Ndi73/AuLDbKCDXU+6+f6CCDXY+Xxe3hg3hPGXBDHJmTHYwkYiIOOTXwE0Ellro/oNoOYG19yRCzjjpCKZNLAKg1dvJH594jbpPVzmcSkTEWX0t9g4GtgGbPOXu9UALkDxgqaLA4/99g42VtQAMT03mG0sX4UpPcziViIhEmqfc7fWUuy/ylLvTgHxgGdAEvOJsstiSkBDP73/yndB2+dsreenZ1+hsaXUwlYiIs/pa7P0JeJPt6+pNAd4vLSlbMiCposBNd5SHHp+431RGz5rhYBoRERkMPOXuak+5+3FPufsMAhOZSQQdedC+HL5oNgB+a7ni0RfY8vZHDqcSEXFOX4u9ScBMT7n7FQBPufstAksvXDZQwQazD1eu46V3VgAQH2f4TskSXMPSHU4lIiKRUlpS1pcLtLXYW4QZY7jhl+eS5ArMP7dicw1/eeRZWqtqHE4mIuKMvhZ7GZ5ytzd8h6fc3Qjk9H+kwe/mOx4NPT50ejF7LZrrYBoREXHA404HkB2bVDyKi88tCW3f9OxbfPCf17B+v4OpRESc0ddF1dNKS8oeBZ4EGoFhwDEEppeOKc0tbXieeCm0feayg0jK1NTOIiIx5qDSkrLG3RyTEpEk8gXf/87XeOSfr/DZ2s20ejv5ledp7p+7F8MnaS1cEYktfe3Zuws4HrgN8ATvjwPuGKBcg9ar//uQdq8PgHEjsjjq+MMcTiQiIg6IA9J3c+trGyv9LMmVSNmvvhva/s8n63jogX/R1eHdxatERIaevvbsXUvg2oPvAEXABuAvwA0DlGvQeunFt0OPF0wfR0p2loNpRETEISsJrKW3Mwb4eYSyyA4cMHcap594GHc/8iwAv/3Hixy2ZC4TDtzf4WQiIpHTp2LPU+72A9cEbzHt1eDELAAH7LeXg0lERMRBV3vK3X/b1QGlJWXxkQojO/a7H5/BP595ndqGZiobWvjtbeXcPHkCqXkxOeWAiMSgPg0xKS0pu6S0pOzN7qUWSkvKDi0tKXu6tKRszMDGG1y8rW28u7oitL34EP11UEQkFu2u0Asec2ckssjO5QwfxrU/Pyu0fd/rH/HY/f/E7+tyMJWISOT09XqCZUCpp9z9HICn3P0f4HLgxoEKNhi9/dp7NHd0ApCTkcqkSWMdTiQiIiK7curxSzh84SwArIXL7n2KtW+853AqEZHI6PPF455y96pe268D2f2eaBB7+ZV3Q4/n7TURY4yDaURERGR3jDH85ZofMDwjMIH4loZmfnz9/bTX1TucTERk4PV1gpYxpSVlFwFP0HPphcKBCjbYdHk7+d9Hq0PbC+fv7WAaERER6atR+Tnc/JsLOPWiwNQDy9/9jHv+8jBnXXImJl6TporI0NXX/+GeAcoIzD62JXh/PfDUAOUadFo2b+X99ZWh7QMXqNgTERGJFictXcQ3jl4Y2v7FfU+z4tW3d/EKEZHo19di71Lg3wSmku6+PQE8N0C5Bp11n6xifW1g/VxXQjyzZ0xwOJGIiIh8GTf95gIKcjIB2NbazkW/v5v2bQ0OpxIRGTh9KvY85e5tnnL3UmAUMA8oAP4O/GEAsw0afp+P1978OLQ9a/p4kpNcDiYSERGRL2t4Zjp3XndxaPuFTzdw3TV34e/S7JwiMjT1eaB6aUmZAWYA3yMwjNMDjB6gXINKS2U1767bEtpeuN8MB9OIiIjIV3XYotmc+80jQ9u/e/R5nv2/px1MJCIycHY7QUtpSdls4FTgZAI9ehAYxukFOgcu2uDRXLGF9zZsDW3PnzPNwTQiIiKyJ37/i7N4+c2P+GjNJjq7/Jx9zT28NHMyRdN0iYaIDC077dkrLSn7WWlJ2SfAW8APCQzh7ACqgROBXOCaSIR0ku3yU7d+Mx9vqg7tmz97qoOJREREZE+kJCfx0O2Xk5GSBMDm+mbOvPQGvC2tDicTEelfuxrG+Wugu6p5BTibQM/eCk+5+1FPubvZU+7+9UAHjDR/VxcdDU2hW+OGTazYWEmHLzCev7gwn4K8mFpeUEREZMiZWDyK26++MLT9/Ip1XPmrW7HWOphKRKR/7arYOwy4C/gIeBR4zFPubgCG7P+Cnc2trFn+DOv+9VzoVvm/d3sM4TxgXw3hFBERGQpOWHoQ55+8/fq9P/zjBR5/+N8OJhIR6V87LfY85e7/esrd3wX2AzYAt5SWlD0K5JWWlCUDlJaUnRGZmJHRVLGZrg7vF/aHF3sLdL2eiIjIkHHNL89l7pSxAHT5Ledc9Vc++/Azh1OJiPSP3U7Q4il3ewn07D1aWlKWReB6vUdLS8o6gH2Avw5sxMjxd/pCj+OTXMQnubDW8t7GqtB+Tc4iIiIydCQmJvDQX37J3GMuoK65lZrmNk48/3e88GgZw3OHOx1PRGSP9HnpBQBPubveU+6+w1PuPga4EBhSVzL7fdvX2cmeOpFxxxxC/KyZVDW0AJCelsJek8c4FU9EREQGQOGoEdx3wyUkxAV+Lfp0cw2nnH0lnd4vjvYREYkmX6rYC+cpd28EFvVjlojy+br4bM0mNlXWsq2hmc5OHzZsUVWTEA/Aa++sCO2bN2sK8fHxEc8qIiIiA+vQxftR9qPTQtv/+WAV7kuvdzCRiMie2+0wzl3xlLu39VeQSKuua2D+sZf02OdKiCclMYHE+DjiXS7iEuJpamkLPa8lF0RERIauc886gRWfr+fWR58D4LbHX2bqxDGcf0Gpw8lERL6aPSr2ollLa/sX9nl9XXhDQznbvvC8JmcREREZ2v5w1YV8vm4zz76zEoAf/dHDxEljOPLIhQ4nExH58r7yMM5o1+X3M7G4gIK84WRmpBEXt+sfxaL9ZnDIAftEKJ2IiIg4ISEhgQf+8kumjh4BgM9vOeXiMt55d8VuXikiMvjEbM/elPGFvPFkGQDprnSstax++iVqKirp7Oqi4IB9SR2RA0B8fBz5mpFLREQkJgzLzOAfd/6SRSddSnVTK03tXpZ991e8+Mh1jB9X6HQ8EZE+i1jPnjHmKGPMSmPMKmPMZTt4/nRjTLUx5r3g7buRyhZ8fxLjDMPTkskblsbokbmMys9hVH6OCj0REZEYM37iWB695TLSk1wAVDW0cMy3fkZ1db3DyURE+i4ixZ4xJh64BTgamA6UGmOm7+DQB621s4K3OyKRLZzft32dPaNZN0VERGLavANmc+9V55MYH/h1ac3WOo791k9obvnidf0iIoNRpIZx7g+sstauATDGPAAcB3yyJyf1Wz/N3uav/Prer+0KK/babQe+PTj3QNuTz+00ZY+8aM0Nyu6UaM4u0p+WHncIt1TVcs4192AtvLu6gpO+cznL//47EhNj9moYEYkSkRrGORrYGLZdEdzX2wnGmA+MMY8YY4p2dCJjzNnGmLeMMW/V1NT0a8ge6+ypZ09ERESAb3/3RH717WND2/95+1PcP/2jg4lERPpmMP1J6p+Ax1rbYYw5B/gbcEjvg6y1twO3A8ydO9emu9L3+I1D5+jyb9+XkkG8y7XH5x5o/fH5naLskRetuUHZnRLN2UX6izGGH//0LCqr6/jTE68CcFv5cxyw7zRKTz7a4XQiIjsXqZ69TUB4T11hcF+ItbbWWtsR3LwD2DdC2UL86tkTERGRHTBxcfzh2ks4dOaE0L7zfv0XPvrwMwdTiYjsWqSKvTeBScaYccYYF3AysDz8AGNMQdjmMiCiC9pYvx/8Nhgm8J+6iIiISLd4VyL3/OnnFGYPA6Clo5OTz7uKxm2NDicTEdmxiFQ01lofcAHwFIEi7iFr7cfGmCuNMcuCh11ojPnYGPM+cCFweiSydevdq2eMieTbi4iISBQYUZDL/Tf8CFdCYATQZ1tqOev7V/WY5E1EZLCIWPeVtfZJa+1ka+0Ea+1vg/sut9YuDz7+ibV2hrV2H2vtEmvtp5HKBmB924u9OA3hFBERkZ2Yf8Asfv+DU0Lb5a9/zA2/vwtrrYOpRES+SGMVg8J79uISBtO8NSIiIjLYnHfOSXzj0P1C27+890k+eesjBxOJiHyRir2g8J49Tc4iIiIiu2KM4bbrL2VKUR4AXl8XV/zh3h5/PBYRcZqKvaCePXsq9kRERGTX0lKTufm33w9tL397Ba/++2UHE4mI9KRiL0g9eyIiIvJlHXzALA7bbzoA1sKVf36YztY2h1OJiASo2AtSz56IiIh8FVf/4hy6J/F+fsV6/vXwU84GEhEJUrEXZLWguoiIiHwFe08fz0mHzw9tX3XP47RW1zqYSEQkQMVekF9LL4iIiMhXdOVlZ5IYH/i16p31W3nonuVaikFEHKdiL6hHz56GcYqISIQYY+4yxlQZY3Y4b78J+KMxZpUx5gNjzJxIZ5TdGz9mJN856fDQ9rWPPk/96vUOJhIRUbEXop49ERFxyN3AUbt4/mhgUvB2NnBrBDLJV/DzH5xKarILgM+31vHXux+jy9vpcCoRiWUq9oLUsyciIk6w1r4I1O3ikOOAe2zA60CWMaYgMunky8jPHc6Fpx8X2v7DE6/y2StvO5hIRGJdgtMBBgv17ImIyCA1GtgYtl0R3LdlVy/yWz/N3uav/KZ78lqnOZn93DOO4o4H/kVNfTPVTa1c9Ie/45k2hqTsYbt9rX7mzlD2yIvW3BB92dWzF6SePRERiXbGmLONMW8ZY96qqalxOk5MGpaRyk1XfS+0/fyn67nt5oc1WYuIOEI9e0Hq2RMRkUFqE1AUtl0Y3PcF1trbgdsB5s6da9Nd6Xv85v1xDqc4lf2EIw7inG++z20PPg3A1f/3HIcdOp/9l8zr0+v1M3eGskdetOaG6Mmunr0gf5cv9Fjr7ImIyCCyHDgtOCvnfKDBWrvLIZzivGt+cTZTx4wEoL2zizMvv5WWpuga/iUi0U/FXlD4MM44DeMUEZEIMcZ4gNeAKcaYCmPMmcaYc40x5wYPeRJYA6wC/gKc51BU+RJSkpP4+02X4Qr+TrFySy2X/uwmh1OJSKzRMM6g8GGc6tkTEZFIsdaW7uZ5C5wfoTjSj/aeMYErzzuJy/74AAB/efJVDj/0vxx/3CEOJxORWKGevSD17ImIiEh/++H3T2Hx3hND22f89GY++HiVg4lEJJao2AtSz56IiIj0N2MMf73hUgqyApM5tHR0UnLmr6iqqXc4mYjEAhV7QerZExERkYEwekwBD5VdTKorEYCNNfWUnPELOjo6HU4mIkOdir0g9eyJiIjIQJl30FxucZ+KMYHtN1es47vua7X+nogMKBV7QerZExERkYF0yndKuOyE7ZOzPPjUa/zuJo+DiURkqFOxF+QPL/biNUmpiIiI9C8TF8fPLz+XE/afHtp3xR89/PWhpx1MJSJDmYo9wFqL7TGMUz8WERER6X+JqSnceq2b/cYVhPZ972c388iTLzmYSkSGKlU1gPX7Q49NXBwmTj8WERERGRhZo/P5+9XfZ2pBDgB+aznth9fxzAvvOpxMRIYaVTXQq1dP1+uJiIjIwCqesxcPXHkO40ZkAeDr8nPaRWW8/MYnDicTkaFExR69rtfT5CwiIiIywIwxTDt4Pvf95NuMDq7B1+H1Ufq9a3jjvZUOpxORoULFHurZExERkcgzcYY5xyzhHvcpjMhIBaClrYNjvv0LXn1bPXwisudU7KGePREREXFGXEI8B5xwJH+94ASGpyYD0NjSxtLTL+f51z5wOJ2IRDsVe/RcY089eyIiIhJJ8UkuFn9zKXedezzZaSlAoIdv2Xev4KkX3nY2nIhENRV7gN8Xvsaeij0RERGJLFd6GgecdCh/+97x5AWHdLZ3dHLCub9h+TOvO5xORKKVij169expGKeIiIg4IDEznf2/voS/nVdCQWZg0hZvp4+TL7ia+/7xnMPpRCQaqdhDPXsiIiIyOCRmZbDwxKO553slFGUPA8DX1cUZl5RxzZ8fxlrrcEIRiSYq9lDPnoiIiAweycMzmff1I7jne8czMW94aP/Pr7uHi674M11hv7eIiOyKij3UsyciIiKDS3J2FnOWHca955Ww37iC0P4/3/ck37zgatraOxxMJyLRQsUe4O/yhR6rZ09EREQGg5Sc4cxYegh/Oed4jp45IbR/+TOvc8S3fkZl9TYH04lINFCxR89F1dWzJyIiIoNF8vBMJh55EH84fSlnHLh3aP//3lvJASVu3v1olYPpRGSwU7FHz0XV1bMnIiIig4krI53iIxZx2TcP47KlC4gzBoCKyhoWn/xjHn7iJYcTishgpWKPnhO0qGdPREREBpuElGTGHLKQM49dxC3/70gykl0AtLV7OfWia/jl9X/H7/c7nFJEBhsVe/ScoMWo2BMREZFBKD7JRdGSBRy1ZD/uO+c4inMzQ8/97pYH+fo5v2FbQ7ODCUVksFGxR6+ePQ3jFBERkUEqLiGBUQfux5xF+3LfOcexcFJh6Lknn3uTecddxNsffu5gQhEZTCJW7BljjjLGrDTGrDLGXLaL404wxlhjzNxIZVPPnoiIiEQLYwx5s2YwadF+3HLaUT0mbllXUcXB37iU2+57Uguwi0hkij1jTDxwC3A0MB0oNcZM38FxGcBFwP8ikaubevZEREQk2mRNLGbskgVc8rWF3HDKYaQnJQLg7fTx/V/eymnu62hsanU4pYg4KSFC77M/sMpauwbAGPMAcBzwSa/jfg38HvhRX07qt36avV99bHr3a32d3tC+duvF7sE5I2VPPrfTlD3yojU3KLtTojm7SCxJG5nHmMMXcXRKMpPzc3A/8CyfbqkF4MF/vsgb763krmt/yMK5MxxOKiJOiNQwztHAxrDtiuC+EGPMHKDIWvvErk5kjDnbGPOWMeatmpqafglnu7bPXqVhnCIiIhJNkoZlMPbwg5i2zxTuPXsZJ8ydGnpu7catHFL6E3527d/wejsdTCkiTohUz94uGWPigDLg9N0da629HbgdYO7cuTbdlb7H79/q3z6mPT05A1c/nDNS+uPzO0XZIy9ac4OyOyWas4vEkrjEBEYtnEtyThZXuBKYN34Uv17+Mk3tXqy1XHvbIzz14tvc/YeL2WvyWKfjikiERKpnbxNQFLZdGNzXLQPYC3jeGLMOmA8sj9QkLVYTtIiIiEiUM8aQM20ShQcv4Ni503j0+ycwb/yo0PMfrFjL/ON+wNV/eojOTp+DSUUkUiJV7L0JTDLGjDPGuICTgeXdT1prG6y1udbaYmttMfA6sMxa+1Ykwvk1QYuIiIgMEWkjRzD2qIMZN2kst59+DJctXUBS8Pcbb6ePy8vu5cATL+GDT9c6nFREBlpEij1rrQ+4AHgKWAE8ZK392BhzpTFmWSQy7Ip69kRERGQoSUxNoWjxAeTtM41TD9iLB88rYa/RI0LPv/vxauYf/0OuvPF+XcsnMoRFbJ09a+2T1trJ1toJ1trfBvddbq1dvoNjF0eqV89ai/WHT9CideZFREQk+pk4Q86MyRQdspApxaO49+xluI/cH1ewl8/n6+I3N3nYb9lFvPi/Dx1OKyIDIeYrm/A19kx8PMYYB9OIiIiI9K/UETkUH3kww8cXccaifXjk/K8za0x+6PkVqzZy2Kk/5Ts/up6q2noHk4pIf1Ox59veq6fr9URERGQoik9yMWrBvoxaOJeJo/O4+7vH8uNjFpDqSgwd8/fy/zLziO/xF8+/6Ar7Y7iIRC8Ve126Xk9ERERiQ0bRKIqPWUxm0Si+dcBePHbRiRw+Y1zo+W0NzZz/iz8x7/gf8p9X3nMwqYj0BxV7XerZExERkdiRkJzMqAP3Y+S82YwekU1Z6WH86bSjKByeETrmgxVrOfrbv2DZmVfw8WfrHUwrIntCxZ569kRERCTGGGPIHFdE8TFLGFZcyKLJRZRfeCLnHTKHlMSE0HH/fuFt9j32Qr73s5vZUlXnYGIR+SpU7IX37KnYExERkRiSkJxEwfw5FC6eT0bWML53yL7884ffoGTOZLrnrPP7/dz54FNMO/Rsrrjh7zQ1tzobWkT6TMVeeM+ehnGKiIhIDEobmUfx0YvJnj6JkVkZXPn1g3n4/K+zYMLo0DGtbR1cdfODTDv0HP789yfo7PQ5mFhE+kLFnk89eyIiIiJxCQmM2HsaxUcvJq0gjykjc7jt9KP587ePYsrI7NBxVbX1XHjFn5lx+Dnc+cBTWpRdZBBTsaeePREREZEQV0Y6hQfPZ/Si/XFlpLFwUhEPnlfCb084mJGZaaHj1lVU8b2f38y0Q8/mrgeeoUNFn8igo2IvrNhTz56IiIhIQProkRQfvYTcfaaRmORi2ezJPP6Db+A+cn+yUpNCx23cUsMlV97FnCMu4sa7/kFLa7uDqUUknIq9sAla1LMnIiIisl1cfDw50yYx7thDGT55PEmuRM5YtA9PXVyK+8j9yU5LDh27pWobP7rqTiYe/B1+c5OHuvomB5OLCKjYU8+eiIiIyG4kJCWRN2cvxh2zhPSiAlKTAkXfvy4+mR8dPZ/c9JTQsbXbmrjyxvuZePCZ/Ph3d7Jhc5WDyUVim4q9sAlatM6eiIiIyM65MtIZvXA/xh5xEGkFeaS6Ejlt4Uz+dfHJ/PxrCxkdtjB7c0sb19/5D6YsOYtv/eBa3vrgMweTi8QmFXvhPXsaxikiIiKyW8nZWRQePJ+iQxeSMiKH5MQEvjlvOo//4BtcdeJiJuYNDx3b1eXnocdf5ICvX8whpZfx6L9fwefr2sXZRaS/JDgdwGk9rtlTz56IiIhIn6WOyKHokAOo3bSRpk/XQU09X5s1iaV7T+SlzzZwzysf8sbaLaHjX37zY15+82MKR+by3dKjOPObR5CfO3znbyAieyTme/b86tkTERER+cqMMSTnZTPioDkUHbKQ1Pxc4uIMB08dy51nHstD55Vw7D4TSYjb/mtnRWUNV1z/d8Yv+g6nua/judfex+/37+JdROSriPlir+cELTHf0SkiIiLylaXm5VC05ADGHHYgaaPyAZg2KpffnbSEf19yMmcvnk1O2GQunZ0+Hlj+Akf+v58z9ZCzuPLG+1lXsdWp+CJDjoo9n5ZeEBEREelPKbnZFB40j+Kjl5A5bgwmLo78YWl8/7C5PH1JKVeftIRZY/N7vGZdRRW/ucnD5MXf5Yhv/Yx7y/+rNftE9lDMd2Vp6QURERGRgZGUmcHIebPI3Xsq2z5bQ/2q9biApftMZOk+E/l0Sy2Pvv0p//pwDfUt2wu751//gOdf/4CLrvgzJxy9kNO+figL504nLi7m+ylEvhQVe1pUXURERGRAJaQkM2Kf6eTMmEzjugq2fb4Wb0MTUwty+OmxC7nkqPk89+l6Hnvvc175bCN+vwUCyzf87ZFn+dsjz1JUkMuJxxzISUsXse/MSRhjHP5UIoOfij317ImIiIhERFxCAlkTi8mcMJbWqhrqP1tL8+ZKXAnxHLnXeI7cazxVjS08/v4qlr+/mtWVtaHXbtxSw/V3/oPr7/wH44tGcsIxB/LNrx3EzCnFKvxEdkLFXlixp6UXRERERAaeMYa0/BGk5Y+gs7WNhtXraVizAV9bO3nD0vjOon0448C9+bCimsfe+5ynPlpLQ0tb6PVrNlZy7W2PcO1tjzB1QhHfOHYRJy1dxJTxhQ5+KpHBR8Ve2AQtWnpBREREJLISU1PInTmVnBmTad68lYbV62mprMJg2Lsoj72L8rjs6Pm8tnoTT69Yz38/WUdT2MQtn67eyJU33s+VN97P3tPGseyw+Sw7fD77TBunHj+JeSr21LMnIiIOMsYcBdwIxAN3WGuv7vX86cC1wKbgrputtXdENKRIBJi4ODIKC8goLMDX1k7Duo00rtmIt6mZxIR4DpoyhoOmjKFj6QJeWVXBMys38p8P19DW4Q2d44MVa/lgxVp+c5OHsaPzWHb4fJYesj8Hzp2Oy5Xo4KcTcYaKvS717ImIiDOMMfHALcDhQAXwpjFmubX2k16HPmitvSDiAUUckpCSTM60SWRPnUh77TYa11XQuGETfm8nSYkJHDKtmEOmFdN6zHxeXLmBpz/dyIufrKWj0xc6x/pNVdx093Juuns56WkpHLpwFscsnsuRB+/LqPwcBz+dSOTEdLFn/X6wgdmeMAaj6XxFRCSy9gdWWWvXABhjHgCOA3oXeyIxyRhDSm42KbnZ5M3ei+YtW2meRjGEAAAdVUlEQVRcX0HLpq1Yv59UVyJHzZzAUTMn0NJxAC9/XsELqzfz/Mdrewz1bG5p47GnX+Oxp18DYO9p4zhi0RyOWDSHvWeOweWK6V+JZQiL6W+2evVERMRho4GNYdsVwLwdHHeCMeYg4DPgh9bajTs4pge/9dPsbf7KwfbktU6L1uzRmhsil93kZZCZN42MfSbStqWGtooqOqrqwFrSklyhGT07ly7gzXVbeGH1Jl5auZGNVdt6nKd7uOd1t/8faalJLNhvKocfOJslC/dmwtiRUXOtX7R+Z6I1N0Rf9tgu9ny6Xk9ERAa9fwIea22HMeYc4G/AITs60BhzNnA2QNGYosglFImwOFciaWMLSBtbgN/bSdvm6kDhV70NrCUxIZ4DJhZywMRCLjtif9bWNPDy6k28vGYzb322kc6w3wFbWjt49oX3efaF9wEoLMhlycKZLF4wk4MX7EV2VoZTH1Nkj8V2sac19kRExFmbgPCqrJDtE7EAYK2tDdu8A7hmZyez1t4O3A4wd+5cm+5K3+OA/XEOp0Rr9mjNDQ5ld8GwycNh8mS6vF6aN2+luWILLVuqsV1dGGMYPyKL8SOyOG3+DFo7Onlj3Rb+t7GaF1esZ8PW2h6nq9hSw72PPMe9jzyHMYbZMyZw6MJZHHbgLBbMmUZykivyn3E3ovU7E625IXqyx3ixt30Yp9EwThERibw3gUnGmHEEiryTgVPCDzDGFFhrtwQ3lwErIhtRJHrEu1xkFheRWVyE3+ejpbKalk1bad5cSVdw1s7UpEQWTxnD4iljuPTQOWyobeT1jVt5Y30Vr36yluawa/2stbzz0Sre+WgV1972CEmuRBbMmcbB82eyZMHezJ05SbN8yqAW48WeevZERMQ51lqfMeYC4CkCSy/cZa392BhzJfCWtXY5cKExZhngA+qA0x0LLBJF4hISQks5WGtpr6uneVMlLZu30lHfCAQmgBmbm8nY3Ey+OXsynV87gI82VfPG5lpe+7yCdz/fgN9vQ+fs8Hby/Osf8PzrH/CrG+4jNSWJ+bOnsmj/vVi03wz2nzVlUPb8SeyK8WJPPXsiIuIsa+2TwJO99l0e9vgnwE8inUtkKDHGkJIznJSc4YzYexqdrW20bN5Ky5YqWiqrQr8TJsbHMXtMPrPH5HPO/Ok0tnXwdkU1b1TU8PpnG1i1cWuP87a2dfDfV9/nv68GrvdzJSYwd+9JzJ8zjQVzprJgzjTycrIi/nlFusV2sedTz56IiIhIrElMTSFrYjFZE4tpamuko7Yef3UTLZVVeBuaQscNS0liyaRClkwqhCWz2NrYwrtbtvFWRRWvf7qe9VtqepzX2+nj1bdX8Orb20dbTxhTwLzZU5g3ayrzZk9h5pRiEhNj+ldwiaCY/qapZ09EREQktpn4OJLzskkvHAPMoLOljZbKKlorq2nZWo3f2xk6Nn9YGkcNS+OoKYVw6BwqG1p4v7qBdzdV88bKDXy+ofIL51+9YQurN2zh/seeByAl2cWcvSay714TmTNzEvvOnMik4lHEab1nGQAxXez5dc2eiIiIiIRJTEsha8JYsiaMxfotHfUNtFRW01pZTVtNHda/vbNgZGYaIzPTOHLiKDh4H2pb21lR18wHW+p4Z80m3lu5no6wYhGgrd3LK299witvfRLaNyw9lZlTi5k5pTh4P44Zk8eQkZ4asc8tQ1NMF3vhE7SoZ09EREREwpk4Q3J2FsnZWeRMn4Tf56Otuo6WrYHir3uil245qckcmJrMgYW5sN9kvL4uVje28VF1PR9sqOLtlevZWFnzhfdpbG79QgEIMLF4FLNnTAjd9p46jhE5mQP6mWVoifFib/tfZtSzJyIiIiK7EpeQQFpBHmkFeQB0dXhpra6ldWsNrVU1Pa73A3AlxDMtO51p2emcNKUQDp9DbbuXzxta+aSqno/WV/LeZ+uprmvc0duxat1mVq3bzMNPvBTalzM8g6kTipg6oYhxxXlMGjeKfSZPZOzoEcTr91npJbaLvR4TtMT0j0JEREREvqT4JFdoeQcAX0cHbdV1wVst7dsawNoer8lJdpGT7GJ+fhbMLMbaedT5/KxtamNVbSMrN9fwydrNrFy7ia6wjolutduadtgLmORKZMLYAqaML2TKhEKmjC9k2sQiJo8bTXpaysD9EGRQi+kKRxO0iIiIiEh/SUhK6lH8+X0+2mvraaupo612G201dT0mfIHAshA5ifHkZKczNzsdJo2Cg/fGl5DI+pYOVlZvY8WmGj5cs4lP126iJWzR93Ad3k4++XwDn3y+4QvPjc7PYdK40UwaN4qJY0cxadxopkwoZHxRvnoDh7gYL/Y0QYuIiIiIDIy4hARS83NJzc8FwFpLZ1MLbXXbaK+tp72uno5tDT0mfemW4OtkQlIcEwpzOKYwB+ZNgbh46ixsaGpjbV0jn22pZs3mGlavr2RrTf1Oc2zaWsumrbU8//oHPfYnuRKZNG400yYWBYeGFjJ1YhGTikdpcfghImLFnjHmKOBGIB64w1p7da/nzwXOB7qAZuBsa+0nXzhRP9IELSIiIiISKcYYXMPScQ1LJ7O4CAiMNOtoaKR9WwPt2+rpqGugo75xhwUg/i6ygex0F7PSc2FMoIiMT3LRkehiY0s76+uaWFu1jc83VfHZusCyD76wS5fCdXg7+WjlOj5aua7H/ri4OMYV5jN5/GgmFo9iwpgCJowN3MaOztM6gVEkIv9Sxph44BbgcKACeNMYs7xXMXe/tfbPweOXAWXAUQOZSxO0iIiIiIiTTHxcaMZPGAuA9fvpaGyiY1sjHfWB4q9jWyNdXu8Oz9HV4SWhw8s4YNzwFBYPT4Epo4hL3A+TlsLW9k4qGlpZX9fAuqptrK6oYuWaCjZvrdvh+fx+f2h9wN7i4+MYM2oE44pGBm/5FBeNpHh0HmML88jLycIY018/HtlDkSrL9wdWWWvXABhjHgCOA0LFnrU2fBqiNKDn1awDIHyCFvXsiYiIiMhgYOLiSM7KJDkrEwj2AFpLV3tHoPBraKSjvom2+np8TS09OjDC+Ts7ob6TXCA3yTCrIAsKsmDWeBLTU+lISGRdfQtraxtYVVnL6k3VrFyzibUVW7F2x7+Kd3X5WbtxK2s3bgXe/8LzKckuxozOY0zBCIpGbb+Ftgty++mnJH0RqWJvNLAxbLsCmNf7IGPM+YAbcAGH7O6kfuun2dv8lUP5fNv/OtJhvZg9OFek7cnndpqyR1605gZld0o0ZxcRGYqMMSSkJJOQkhxa+qHZ24y1lqTOOLwNTXQ0NOFtbMLb2Iy3sRm/z7fjkwWvHYwDxgPjs1M5NDsVphcRn7wAm5xMRUs7Gxta2FDbyPqqOtZuqmbV+s077Q3s1tbuZeXqClaurtjp58jPzWTUyByKCvIYnZ9DQV42BfnZjM7PYfTIHEbn52hB+X4yqAbcWmtvAW4xxpwC/Bz4du9jjDFnA2cDFI0p2rP3C5+NU8M4RURERCTKGGNwpafhSk8jffTI0P5QT2BDE96m5lAB6G1qxtfattPzdbV3QHsHo4BRqQnMS82GomzMfpNJTE+ly+Via5uXzY2tVGxrYkNNPRsr61i/qYr1m7ZS39iyy7zWWiqr66msruedD1fv9LiMtBRGj8ylIG84+SOGMzI3cJ+fm0XhyFwKCwI3TSSza5Eq9jbR3QcdUBjctzMPALfu6Alr7e3A7QBz58616a70rxwqzm/oHsiZlpxO8h6cyyl78vmdpuyRF625QdmdEs3ZRURiWY+ewJEjejzn9/nwNrX06AXsaGyis6llxxPDELiO0NsYGPWRC+TGw965aZCbBtMLSUxNJTE9lVZjqGrpYGtTG5vrm9hS18jGylo2bqlh4+ZqNlfV4d/Je4Rramnj09Ub+XT1xl0eNyI7k4L8bLIzM8jKTAvep5Ofk8XIvOEUjMgO3OdlMyw9NeauJ4xUsfcmMMkYM45AkXcycEr4AcaYSdbaz4ObS4HPGWA9ZuNUz56IiIiIxIC4hASSh2eSPDyzx37rt3S2tG4vApuaA0VhU3Ogx29n/JbO5hY6mwO9ennB28zMJMgcgZk4ksS0FBLTUrEuF5UtLVQ2t9LQ2kVVQwuV2xrZXFXHlq21bKoMLBPR0Ws9wp2prmuguq6hT8cmJ7nIz80iP3c4eblZ5OdmMSInk7ycTEbkZJGXk0lWRhoZ6akMy0glIy2F5CRXVBeIESn2rLU+Y8wFwFMEll64y1r7sTHmSuAta+1y4AJjzGFAJ7CNHQzh7PdcYRO0xGmCFhERERGJYSbO4MpIw5WRFphxI4y/0xco/ppb6GxqwdsUKO68zS27LgQJdLB09yACDAveiAOGJ8HwESTMKCIhNYWElGTiU5Jo8vmpbm6nrrWdmqZWqhtbqKptYHNVHRVbaqjYEugl7NrJ5DQ70t7hDQ43rerzaxITEygYEegZHJWfw4jcDPJHZFGYl8eInCxGZA9jRE4WucOHkZGeMugKw4hds2etfRJ4ste+y8MeXxSpLKH31DV7IiIiIiK7FZeYELZERE9+n4/O5tZAr2BzC50trfha2uhsaaWzuXXnE8WE8bW142tr77EvJ3ibFAdkJRGXN5qEORMCw1OTk8GVQF17J9tavTR2eGlo99LY5mVbYzNba7axpWobldV1VFZtY0t1HW3tO166Ylc6O31s2FzNhs3Vuz02ISGe7MwMhmelk52ZwbD0FNLTUhiWHuglzEhPZXhmOlnD0sgaFrzPTGdcYf6ATUgzqCZoibTwYZzq2RMRERER+fLiEhJIyhpGUtawLzxnrcXv7aSzNVD8+VrbaG1qpKu1Hdveia+17QtF3s74vZ14vZ14G5pC+xKAEcFb9464/HQSxuYSn5xEQnIS8clJxCe5aLdQ19pBTXMrtc1t1Da2UL2tieraeqpqG6iuraehuZWm5laamttobG7F27n7QrWbz9dFVW09VbX1fX4NwH03XspJSxd9qdf0VcwWe9Za9eyJiIiIiAwgYwzxSS7ik1yhawQTg8v7dE8CZrv8dLYFij5fazu+tjZ8re2Bfa3t23v9drL2X2/+Th/ezmZo+uIyQkkERqiOBkiEuNGZxI/PCxaFLuKTtheICUlJdGCpbmylqr6ZyroG1m6ppLq2gYaGNqprG6iqrae6toHa+iZaWvtWtPY2PHPgJkOL3WKvR6EXN+jG14qIiIiIxAITHxdaPmJnupeS8LW14+u+b2vfvi9sf1+LQggUhv5OX2hymR2JA0YGb7NzMogryCYhKSlUxMa7Avc+Y2js8NHQ3kFjm5dWbyct3k6aWttpbmmnobmFhsYWtjU0U9/YQn1j4H5EzheHxvaXGC72NBOniIiIiEg0CF9KYlestXR1eANFYHtH6N7X3k5Xu5eujp77v0xhCIEJHrt8XXTtohcvNAFNd/b4OOJHpBE3OitQGLoSiXMlEu9KJN7lIn1U7pfK8GXEbLHnD5+JU8WeiIiIiEjUM8aQELxWL2k3x3ZfT+jr2F789SgUOzoCBaLXG9jv9cKXqw0D7+P3B3ocd3Jtoiszg6TMjC9/4j6I2WKvR8+eJmcREREREYkp4dcTMmz3xZa1lqaWevzeTpJsYqAA7AgrBns99ns76erw7nSh+m7xrsT++khfELPFXlxSIpl7T8L6ukhO3vn4YBEREREREWMMccEhmCmuvk2qEpgUsitYCHYGCkBvz8eJaQNXi8RssZeQlET6xCJg+0xAIiIiIiIi/cUYg0lIIC4hgUQH+pfiIv+WIiIiIiIiMtBU7ImIiIiIiAxBKvZERERERESGIBV7IiIiIiIiQ5CKPRERERERkSFIxZ6IiIiIiMgQpGJPRERERERkCFKxJyIiIiIiMgSp2BMRERERERmCVOyJiIiIiIgMQSr2REREREREhiAVeyIiIiIiIkOQsdY6neErM8ZUA+v38DS5QE0/xHGCsjsjWrNHa25QdqcMtuxjrbUjnA4RLdRGRm32aM0Nyu6UaM0erblh8GXfafsY1cVefzDGvGWtnet0jq9C2Z0RrdmjNTcou1OiObv0j2j+DkRr9mjNDcrulGjNHq25IbqyaxiniIiIiIjIEKRiT0REREREZAhSsQe3Ox1gDyi7M6I1e7TmBmV3SjRnl/4Rzd+BaM0erblB2Z0SrdmjNTdEUfaYv2ZPRERERERkKFLPnoiIiIiIyBAU08WeMeYoY8xKY8wqY8xlTufZFWPMXcaYKmPMR2H7so0xzxhjPg/eD3cy444YY4qMMc8ZYz4xxnxsjLkouD8asicbY94wxrwfzP6r4P5xxpj/Bb83DxpjXE5n3RljTLwx5l1jzOPB7ajIboxZZ4z50BjznjHmreC+aPjOZBljHjHGfGqMWWGMWRAluacEf9bdt0ZjzA+iIbsMDLWPkaE20jlqHyNPbaQzYrbYM8bEA7cARwPTgVJjzHRnU+3S3cBRvfZdBvzHWjsJ+E9we7DxARdba6cD84Hzgz/naMjeARxird0HmAUcZYyZD/weuN5aOxHYBpzpYMbduQhYEbYdTdmXWGtnhU1tHA3fmRuBf1trpwL7EPjZD/rc1tqVwZ/1LGBfoBUoJwqyS/9T+xhRaiOdo/Yx8tRGOsFaG5M3YAHwVNj2T4CfOJ1rN5mLgY/CtlcCBcHHBcBKpzP24TM8BhwebdmBVOAdYB6BRTQTdvQ9Gkw3oJDAfz6HAI8DJoqyrwNye+0b1N8ZIBNYS/Ba6GjJvYPPcQTwSjRm163fvgNqH537HGojI5NX7WPkc6uNdOgWsz17wGhgY9h2RXBfNMm31m4JPq4E8p0MszvGmGJgNvA/oiR7cJjHe0AV8AywGqi31vqChwzm780NwKWAP7idQ/Rkt8DTxpi3jTFnB/cN9u/MOKAa+GtwaNAdxpg0Bn/u3k4GPMHH0ZZd+ofaRweojYwotY+RpzbSIbFc7A0pNvBnhUE7taoxJh34P+AH1trG8OcGc3ZrbZcNdNsXAvsDUx2O1CfGmGOBKmvt205n+YoOtNbOITCM7HxjzEHhTw7S70wCMAe41Vo7G2ih15COQZo7JHiNyjLg4d7PDfbsIjsTDd9dtZGRo/bRMWojHRLLxd4moChsuzC4L5psNcYUAATvqxzOs0PGmEQCjdh91tpHg7ujIns3a2098ByBoR1ZxpiE4FOD9XuzEFhmjFkHPEBgqMqNREd2rLWbgvdVBMbF78/g/85UABXW2v8Ftx8h0LAN9tzhjgbesdZuDW5HU3bpP2ofI0htZMSpfXSG2kiHxHKx9yYwKTj7kotAt+xyhzN9WcuBbwcff5vAWP9BxRhjgDuBFdbasrCnoiH7CGNMVvBxCoHrKFYQaNBODB42KLNba39irS201hYT+G7/11p7KlGQ3RiTZozJ6H5MYHz8Rwzy74y1thLYaIyZEtx1KPAJgzx3L6VsH54C0ZVd+o/axwhRGxl5ah+doTbSOTG9qLox5hgC47bjgbustb91ONJOGWM8wGIgF9gK/BL4B/AQMAZYD3zDWlvnVMYdMcYcCLwEfMj2sfE/JXBNwmDPvjfwNwLfjzjgIWvtlcaY8QT+GpgNvAt8y1rb4VzSXTPGLAYusdYeGw3ZgxnLg5sJwP3W2t8aY3IY/N+ZWcAdgAtYA5xB8LvDIM4NoV8cNgDjrbUNwX2D/mcuA0PtY2SojXSW2sfIUhvpjJgu9kRERERERIaqWB7GKSIiIiIiMmSp2BMRERERERmCVOyJiIiIiIgMQSr2REREREREhiAVeyIiIiIiIkNQwu4PERm8SkvKTgcuA6YA93vK3aeGPecGLgTagas95e67B+D9fwb8CMgEzhiI9/iSeU4ErgIsgemwL/KUu59wMlO30pKyEuB6YCzwN0+5+3RnE4mIDG1qI7+QR22kxBz17ElUCzYcVwc3TyktKTsj7Lky4G4GqBELvsdvgfcG4txfVmlJmQu4B5gEHAycDTQ7GiqMp9xdTuDfQ0REIkBt5HZqIyVWqWdPhgovgUU6byotKXvNU+7+1OlADigAUgA85e5KoNLZOCIiMkiojVQbKTFKxZ4MFbcCxxMY/vBAaUnZPE+5uyP8gNKSspuAs4AkYAkwCvgTweElwL3AK8C84EsuA04G8oFzgBzAHTz+LE+5++leGY4uLSk7HjgAeAy40FPubistKYsHLgH+H4HhMl2A21PufqW0pOybwO+DuR8DWoD5wEs7GsJRWlJ2KoEhMe1ABvAIgSEpccCDYcc9T2DIzu07OMfC4HsmAenAfcFzFABPAPsAW4PnPhTwERjq8t/g6ycD1wUzA6wDLvGUuz8PPj8dKAOmAysJ/D9zXa+hMsWlJWV/BZYBnwMnesrdFaUlZcOA24L/BhuALOBZT7n7kt6fQ0RE+kxtpNpIiVEaxilDRT1wCoFGYh8C/5H24Cl3f5+wv+R5yt33Eza8xFPu7iLQcHV7DjiXwH/w9wGfEGgIxwB/2EGGFk+5+3jgAeC7wC+C+y8mMIxmOYH/oPOBp0tLyvI95e4H2T5sYwnwQ2AxUNf75MHx/H8HNnvK3fOBu4DLgas85e628OyecvfinTRiY4Fngp9hPoHG6tfAeZ5y9ybgB8FD84BrCQx1mQb8o7SkLKe0pCwD+A+wFDgKODz4+NnSkrL0YEP0H+BI4HhPufvwYObDe0UpCP6Mrgv+TLobqh8GP8dlnnL3YuBoYGHvzyEiIl+K2ki1kRKjVOzJkOEpd78KXBHcPC/4H/+eeAeoCj7O6LU9eQfHv9nrvrthOTN4/3GwsfwUSCXQ8IZ72VPurvKUuzd6yt3uHZz/7F7nfyN4f25pSZnZ3YcJOoXAMJaVwSwfB/ef1eu4ak+5e72n3F0FrCfw+ZcCxwCFQIWn3L0l7PkxweeWAiOBLZ5y9zvBcz0E3N/r/O8G37/7F4tpwfus4P0VpSVllwLDCDRmIiKyB9RG9onaSBlyVOzJUHMV8Hzw8Z0E/oP9Sjzlbh+BGbt2tO3awUvagvetwfuC4H1R8P6nwaEjYwj85z+s1+trdhOpe0hIa6/7VCB3N6/t1p1lTjCLO5il9/8FbWGPwz9P7wzhj8eGnb+x+0lPubvJU+5+g57ag/fdP8+k4P3NBBrXaQSG0XwK/GaXn0hERPpKbeSuqY2UIUfFngwpnnK3HzgVqAWGA6f3OqT7GoX44H1KP75997lSg/dbgvcVwftrg0NHphMYlnFLr9dbdm19r/N337ew+0awW3eWD4NZ9ifwF9jef+EN/7mEf57eGcIfrydwDQGENdKlJWXDS0vK9u1jvjhgX2A2cE1w3/mlJWXD+/h6ERHZCbWRu6U2UoYcFXsy5HjK3ZsJXEwO0HvoRvcMZCNLS8rSgKn9+Nb79bp/IHj/t+D9YQClJWUJQDmB6Z+/jFuD9/v3ur/VU+7eXSPY7UECjfns0pKy7OC+swhc0B5uRGlJ2djSkrI8An+NbCJwYfoTwEagsLSkrCDs+XVhz28GCkpLymYFz3UOfR9mchFwqqfc/R6B6zmaCPxVdNBMjy0iEs3URu6S2kgZcoy1ff3+iww+YQvGJgN/DK4b1P3cjQQWjA0t5FpaUrYP4CEwW9hDwCICfyVbSeCvnbewfaax3wLFwf0QuIC8KGz7EQIXr3cvGHs/kMYXZxpLBH4CfBvYFrzd7yl3/7W0pGxp8D3HEpjd6z5PufviXXzeUuDHbJ9p7GECw3K66DlL2gvBz712B+dYDPyOwEXwnxFoeC70lLubg889F8xSTuCCeF/w+e6ZxiYSuPh+HIG/tK4FLvaUu1cHn58afH46sJpAI3cegQvQwz/rBQSGn0wBGghc7F5JYNhMHZBNoNH9xWBZ9FZEJJqojVQbKaJiT0RCwhqy9Z5yd7GzaURERAYPtZESjTSMU0REREREZAhSsSciAJSWlI0Gbghujgwu6CoiIhLz1EZKtNIwThERERERkSFIPXsiIiIiIiJDkIo9ERERERGRIUjFnoiIiIiIyBCkYk9ERERERGQIUrEnIiIiIiIyBKnYExERERERGYL+PzAWfu8K2mdTAAAAAElFTkSuQmCC\n",
            "text/plain": [
              "<Figure size 1080x432 with 2 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "8CO8w3uTnjxr"
      },
      "source": [
        "model3_params = {'activation 1': 'relu', \n",
        "                 'activation 2': 'relu', \n",
        "                 'activation 3': 'softmax', \n",
        "                 'opt': tf.keras.optimizers.SGD(learning_rate=0.01),\n",
        "                 'loss': 'categorical_crossentropy',\n",
        "                 'epochs': 45, \n",
        "                 'batch_size': 16, \n",
        "                 }"
      ],
      "execution_count": 44,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 1000
        },
        "id": "Rd_DRvv2n3jX",
        "outputId": "d5d45ce2-adc0-4a87-d1e5-bf49e10ebd65"
      },
      "source": [
        "model3_history, model3 = create_model(train, label_train, test, label_test, model3_params)"
      ],
      "execution_count": 45,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Epoch 1/45\n",
            "575/575 [==============================] - 1s 2ms/step - loss: 1.7529 - acc: 0.5126 - val_loss: 1.3868 - val_acc: 0.5283\n",
            "Epoch 2/45\n",
            "575/575 [==============================] - 1s 2ms/step - loss: 1.1958 - acc: 0.5902 - val_loss: 1.0883 - val_acc: 0.6401\n",
            "Epoch 3/45\n",
            "575/575 [==============================] - 1s 2ms/step - loss: 0.9647 - acc: 0.7132 - val_loss: 0.9049 - val_acc: 0.7398\n",
            "Epoch 4/45\n",
            "575/575 [==============================] - 1s 2ms/step - loss: 0.8066 - acc: 0.7781 - val_loss: 0.7607 - val_acc: 0.8007\n",
            "Epoch 5/45\n",
            "575/575 [==============================] - 1s 2ms/step - loss: 0.6704 - acc: 0.8287 - val_loss: 0.6326 - val_acc: 0.8473\n",
            "Epoch 6/45\n",
            "575/575 [==============================] - 1s 2ms/step - loss: 0.5508 - acc: 0.8648 - val_loss: 0.5239 - val_acc: 0.8677\n",
            "Epoch 7/45\n",
            "575/575 [==============================] - 1s 2ms/step - loss: 0.4542 - acc: 0.8863 - val_loss: 0.4400 - val_acc: 0.8960\n",
            "Epoch 8/45\n",
            "575/575 [==============================] - 1s 2ms/step - loss: 0.3819 - acc: 0.9035 - val_loss: 0.3768 - val_acc: 0.9038\n",
            "Epoch 9/45\n",
            "575/575 [==============================] - 1s 2ms/step - loss: 0.3272 - acc: 0.9210 - val_loss: 0.3274 - val_acc: 0.9252\n",
            "Epoch 10/45\n",
            "575/575 [==============================] - 1s 2ms/step - loss: 0.2847 - acc: 0.9345 - val_loss: 0.2883 - val_acc: 0.9386\n",
            "Epoch 11/45\n",
            "575/575 [==============================] - 1s 2ms/step - loss: 0.2517 - acc: 0.9429 - val_loss: 0.2595 - val_acc: 0.9382\n",
            "Epoch 12/45\n",
            "575/575 [==============================] - 1s 2ms/step - loss: 0.2252 - acc: 0.9467 - val_loss: 0.2327 - val_acc: 0.9456\n",
            "Epoch 13/45\n",
            "575/575 [==============================] - 1s 2ms/step - loss: 0.2034 - acc: 0.9516 - val_loss: 0.2119 - val_acc: 0.9465\n",
            "Epoch 14/45\n",
            "575/575 [==============================] - 1s 2ms/step - loss: 0.1852 - acc: 0.9541 - val_loss: 0.1968 - val_acc: 0.9495\n",
            "Epoch 15/45\n",
            "575/575 [==============================] - 1s 2ms/step - loss: 0.1702 - acc: 0.9558 - val_loss: 0.1808 - val_acc: 0.9495\n",
            "Epoch 16/45\n",
            "575/575 [==============================] - 1s 2ms/step - loss: 0.1570 - acc: 0.9602 - val_loss: 0.1677 - val_acc: 0.9556\n",
            "Epoch 17/45\n",
            "575/575 [==============================] - 1s 2ms/step - loss: 0.1464 - acc: 0.9627 - val_loss: 0.1569 - val_acc: 0.9582\n",
            "Epoch 18/45\n",
            "575/575 [==============================] - 1s 2ms/step - loss: 0.1364 - acc: 0.9639 - val_loss: 0.1474 - val_acc: 0.9595\n",
            "Epoch 19/45\n",
            "575/575 [==============================] - 1s 2ms/step - loss: 0.1282 - acc: 0.9653 - val_loss: 0.1393 - val_acc: 0.9608\n",
            "Epoch 20/45\n",
            "575/575 [==============================] - 1s 2ms/step - loss: 0.1206 - acc: 0.9669 - val_loss: 0.1319 - val_acc: 0.9634\n",
            "Epoch 21/45\n",
            "575/575 [==============================] - 1s 2ms/step - loss: 0.1137 - acc: 0.9713 - val_loss: 0.1253 - val_acc: 0.9695\n",
            "Epoch 22/45\n",
            "575/575 [==============================] - 1s 2ms/step - loss: 0.1075 - acc: 0.9730 - val_loss: 0.1188 - val_acc: 0.9700\n",
            "Epoch 23/45\n",
            "575/575 [==============================] - 1s 2ms/step - loss: 0.1017 - acc: 0.9747 - val_loss: 0.1134 - val_acc: 0.9713\n",
            "Epoch 24/45\n",
            "575/575 [==============================] - 1s 2ms/step - loss: 0.0970 - acc: 0.9754 - val_loss: 0.1090 - val_acc: 0.9743\n",
            "Epoch 25/45\n",
            "575/575 [==============================] - 1s 2ms/step - loss: 0.0920 - acc: 0.9788 - val_loss: 0.1040 - val_acc: 0.9752\n",
            "Epoch 26/45\n",
            "575/575 [==============================] - 1s 2ms/step - loss: 0.0877 - acc: 0.9793 - val_loss: 0.1003 - val_acc: 0.9765\n",
            "Epoch 27/45\n",
            "575/575 [==============================] - 1s 2ms/step - loss: 0.0838 - acc: 0.9827 - val_loss: 0.0964 - val_acc: 0.9800\n",
            "Epoch 28/45\n",
            "575/575 [==============================] - 1s 2ms/step - loss: 0.0800 - acc: 0.9851 - val_loss: 0.0926 - val_acc: 0.9804\n",
            "Epoch 29/45\n",
            "575/575 [==============================] - 1s 2ms/step - loss: 0.0764 - acc: 0.9852 - val_loss: 0.0903 - val_acc: 0.9809\n",
            "Epoch 30/45\n",
            "575/575 [==============================] - 1s 2ms/step - loss: 0.0737 - acc: 0.9866 - val_loss: 0.0865 - val_acc: 0.9809\n",
            "Epoch 31/45\n",
            "575/575 [==============================] - 1s 2ms/step - loss: 0.0704 - acc: 0.9872 - val_loss: 0.0845 - val_acc: 0.9826\n",
            "Epoch 32/45\n",
            "575/575 [==============================] - 1s 2ms/step - loss: 0.0679 - acc: 0.9878 - val_loss: 0.0806 - val_acc: 0.9817\n",
            "Epoch 33/45\n",
            "575/575 [==============================] - 1s 2ms/step - loss: 0.0651 - acc: 0.9881 - val_loss: 0.0782 - val_acc: 0.9817\n",
            "Epoch 34/45\n",
            "575/575 [==============================] - 1s 2ms/step - loss: 0.0627 - acc: 0.9881 - val_loss: 0.0764 - val_acc: 0.9813\n",
            "Epoch 35/45\n",
            "575/575 [==============================] - 1s 2ms/step - loss: 0.0602 - acc: 0.9884 - val_loss: 0.0761 - val_acc: 0.9817\n",
            "Epoch 36/45\n",
            "575/575 [==============================] - 1s 2ms/step - loss: 0.0583 - acc: 0.9881 - val_loss: 0.0729 - val_acc: 0.9817\n",
            "Epoch 37/45\n",
            "575/575 [==============================] - 1s 2ms/step - loss: 0.0563 - acc: 0.9881 - val_loss: 0.0728 - val_acc: 0.9822\n",
            "Epoch 38/45\n",
            "575/575 [==============================] - 1s 2ms/step - loss: 0.0545 - acc: 0.9892 - val_loss: 0.0691 - val_acc: 0.9822\n",
            "Epoch 39/45\n",
            "575/575 [==============================] - 1s 2ms/step - loss: 0.0526 - acc: 0.9889 - val_loss: 0.0687 - val_acc: 0.9817\n",
            "Epoch 40/45\n",
            "575/575 [==============================] - 1s 2ms/step - loss: 0.0512 - acc: 0.9898 - val_loss: 0.0670 - val_acc: 0.9822\n",
            "Epoch 41/45\n",
            "575/575 [==============================] - 1s 2ms/step - loss: 0.0497 - acc: 0.9902 - val_loss: 0.0656 - val_acc: 0.9822\n",
            "Epoch 42/45\n",
            "575/575 [==============================] - 1s 2ms/step - loss: 0.0481 - acc: 0.9904 - val_loss: 0.0641 - val_acc: 0.9817\n",
            "Epoch 43/45\n",
            "575/575 [==============================] - 1s 2ms/step - loss: 0.0466 - acc: 0.9906 - val_loss: 0.0629 - val_acc: 0.9822\n",
            "Epoch 44/45\n",
            "575/575 [==============================] - 1s 2ms/step - loss: 0.0456 - acc: 0.9911 - val_loss: 0.0637 - val_acc: 0.9822\n",
            "Epoch 45/45\n",
            "575/575 [==============================] - 1s 2ms/step - loss: 0.0442 - acc: 0.9906 - val_loss: 0.0632 - val_acc: 0.9830\n",
            "<bound method Model.summary of <tensorflow.python.keras.engine.sequential.Sequential object at 0x7fd8b0293240>>\n",
            "Test Accuracy:\n",
            "6461/6461 [==============================] - 10s 2ms/step - loss: 0.0641 - acc: 0.9841\n",
            "Random Sample Accuracy:\n",
            "94/94 [==============================] - 0s 2ms/step - loss: 1.4451 - acc: 0.6513\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 1080x432 with 2 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "Tr_uT5YW6wBu"
      },
      "source": [
        "model4_params = {'activation 1': 'relu', \n",
        "                 'activation 2': 'relu', \n",
        "                 'activation 3': 'softmax', \n",
        "                 'opt': tf.keras.optimizers.SGD(learning_rate=0.001),\n",
        "                 'loss': 'categorical_crossentropy',\n",
        "                 'epochs': 250, \n",
        "                 'batch_size': 50, \n",
        "                 }"
      ],
      "execution_count": 216,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 1000
        },
        "id": "nKzymFb26_kD",
        "outputId": "afc8f32c-8726-4a94-aa41-51e14393de53"
      },
      "source": [
        "model4_history, model4 = create_model(train, label_train, test, label_test, model4_params)"
      ],
      "execution_count": 217,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Epoch 1/250\n",
            "184/184 [==============================] - 1s 3ms/step - loss: 2.5759 - acc: 0.4680 - val_loss: 2.5382 - val_acc: 0.5109\n",
            "Epoch 2/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 2.4958 - acc: 0.5215 - val_loss: 2.4597 - val_acc: 0.5109\n",
            "Epoch 3/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 2.4155 - acc: 0.5215 - val_loss: 2.3818 - val_acc: 0.5109\n",
            "Epoch 4/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 2.3352 - acc: 0.5215 - val_loss: 2.3019 - val_acc: 0.5109\n",
            "Epoch 5/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 2.2511 - acc: 0.5215 - val_loss: 2.2175 - val_acc: 0.5109\n",
            "Epoch 6/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 2.1621 - acc: 0.5215 - val_loss: 2.1282 - val_acc: 0.5109\n",
            "Epoch 7/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 2.0686 - acc: 0.5215 - val_loss: 2.0364 - val_acc: 0.5109\n",
            "Epoch 8/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 1.9762 - acc: 0.5215 - val_loss: 1.9503 - val_acc: 0.5109\n",
            "Epoch 9/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 1.8927 - acc: 0.5215 - val_loss: 1.8752 - val_acc: 0.5109\n",
            "Epoch 10/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 1.8225 - acc: 0.5215 - val_loss: 1.8147 - val_acc: 0.5109\n",
            "Epoch 11/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 1.7671 - acc: 0.5215 - val_loss: 1.7677 - val_acc: 0.5109\n",
            "Epoch 12/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 1.7243 - acc: 0.5215 - val_loss: 1.7314 - val_acc: 0.5109\n",
            "Epoch 13/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 1.6906 - acc: 0.5215 - val_loss: 1.7021 - val_acc: 0.5109\n",
            "Epoch 14/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 1.6630 - acc: 0.5215 - val_loss: 1.6774 - val_acc: 0.5109\n",
            "Epoch 15/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 1.6392 - acc: 0.5215 - val_loss: 1.6557 - val_acc: 0.5109\n",
            "Epoch 16/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 1.6180 - acc: 0.5215 - val_loss: 1.6359 - val_acc: 0.5109\n",
            "Epoch 17/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 1.5986 - acc: 0.5215 - val_loss: 1.6178 - val_acc: 0.5109\n",
            "Epoch 18/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 1.5807 - acc: 0.5215 - val_loss: 1.6008 - val_acc: 0.5109\n",
            "Epoch 19/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 1.5639 - acc: 0.5215 - val_loss: 1.5847 - val_acc: 0.5109\n",
            "Epoch 20/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 1.5479 - acc: 0.5215 - val_loss: 1.5694 - val_acc: 0.5109\n",
            "Epoch 21/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 1.5328 - acc: 0.5215 - val_loss: 1.5549 - val_acc: 0.5109\n",
            "Epoch 22/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 1.5183 - acc: 0.5215 - val_loss: 1.5410 - val_acc: 0.5109\n",
            "Epoch 23/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 1.5045 - acc: 0.5215 - val_loss: 1.5275 - val_acc: 0.5109\n",
            "Epoch 24/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 1.4910 - acc: 0.5215 - val_loss: 1.5144 - val_acc: 0.5109\n",
            "Epoch 25/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 1.4780 - acc: 0.5215 - val_loss: 1.5018 - val_acc: 0.5109\n",
            "Epoch 26/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 1.4653 - acc: 0.5215 - val_loss: 1.4894 - val_acc: 0.5109\n",
            "Epoch 27/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 1.4530 - acc: 0.5215 - val_loss: 1.4774 - val_acc: 0.5109\n",
            "Epoch 28/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 1.4410 - acc: 0.5215 - val_loss: 1.4657 - val_acc: 0.5109\n",
            "Epoch 29/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 1.4293 - acc: 0.5215 - val_loss: 1.4542 - val_acc: 0.5109\n",
            "Epoch 30/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 1.4179 - acc: 0.5215 - val_loss: 1.4429 - val_acc: 0.5109\n",
            "Epoch 31/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 1.4066 - acc: 0.5215 - val_loss: 1.4317 - val_acc: 0.5109\n",
            "Epoch 32/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 1.3954 - acc: 0.5215 - val_loss: 1.4206 - val_acc: 0.5109\n",
            "Epoch 33/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 1.3844 - acc: 0.5215 - val_loss: 1.4099 - val_acc: 0.5109\n",
            "Epoch 34/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 1.3739 - acc: 0.5215 - val_loss: 1.3995 - val_acc: 0.5109\n",
            "Epoch 35/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 1.3636 - acc: 0.5215 - val_loss: 1.3893 - val_acc: 0.5109\n",
            "Epoch 36/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 1.3534 - acc: 0.5215 - val_loss: 1.3793 - val_acc: 0.5109\n",
            "Epoch 37/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 1.3435 - acc: 0.5215 - val_loss: 1.3694 - val_acc: 0.5109\n",
            "Epoch 38/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 1.3338 - acc: 0.5215 - val_loss: 1.3599 - val_acc: 0.5109\n",
            "Epoch 39/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 1.3243 - acc: 0.5215 - val_loss: 1.3505 - val_acc: 0.5109\n",
            "Epoch 40/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 1.3151 - acc: 0.5215 - val_loss: 1.3414 - val_acc: 0.5109\n",
            "Epoch 41/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 1.3060 - acc: 0.5215 - val_loss: 1.3324 - val_acc: 0.5109\n",
            "Epoch 42/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 1.2971 - acc: 0.5215 - val_loss: 1.3236 - val_acc: 0.5109\n",
            "Epoch 43/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 1.2884 - acc: 0.5215 - val_loss: 1.3150 - val_acc: 0.5109\n",
            "Epoch 44/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 1.2799 - acc: 0.5312 - val_loss: 1.3066 - val_acc: 0.5335\n",
            "Epoch 45/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 1.2715 - acc: 0.5409 - val_loss: 1.2984 - val_acc: 0.5335\n",
            "Epoch 46/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 1.2634 - acc: 0.5409 - val_loss: 1.2903 - val_acc: 0.5335\n",
            "Epoch 47/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 1.2554 - acc: 0.5409 - val_loss: 1.2824 - val_acc: 0.5335\n",
            "Epoch 48/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 1.2476 - acc: 0.5409 - val_loss: 1.2747 - val_acc: 0.5335\n",
            "Epoch 49/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 1.2399 - acc: 0.5404 - val_loss: 1.2671 - val_acc: 0.5361\n",
            "Epoch 50/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 1.2324 - acc: 0.5521 - val_loss: 1.2597 - val_acc: 0.5492\n",
            "Epoch 51/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 1.2251 - acc: 0.5599 - val_loss: 1.2524 - val_acc: 0.5500\n",
            "Epoch 52/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 1.2178 - acc: 0.5599 - val_loss: 1.2452 - val_acc: 0.5500\n",
            "Epoch 53/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 1.2108 - acc: 0.5704 - val_loss: 1.2382 - val_acc: 0.5714\n",
            "Epoch 54/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 1.2038 - acc: 0.5870 - val_loss: 1.2313 - val_acc: 0.5788\n",
            "Epoch 55/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 1.1970 - acc: 0.5941 - val_loss: 1.2245 - val_acc: 0.5783\n",
            "Epoch 56/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 1.1903 - acc: 0.5947 - val_loss: 1.2179 - val_acc: 0.5783\n",
            "Epoch 57/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 1.1838 - acc: 0.5947 - val_loss: 1.2114 - val_acc: 0.5818\n",
            "Epoch 58/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 1.1774 - acc: 0.5981 - val_loss: 1.2050 - val_acc: 0.5831\n",
            "Epoch 59/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 1.1711 - acc: 0.6068 - val_loss: 1.1987 - val_acc: 0.5953\n",
            "Epoch 60/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 1.1649 - acc: 0.6254 - val_loss: 1.1926 - val_acc: 0.6158\n",
            "Epoch 61/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 1.1588 - acc: 0.6295 - val_loss: 1.1865 - val_acc: 0.6158\n",
            "Epoch 62/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 1.1529 - acc: 0.6322 - val_loss: 1.1806 - val_acc: 0.6201\n",
            "Epoch 63/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 1.1470 - acc: 0.6343 - val_loss: 1.1748 - val_acc: 0.6227\n",
            "Epoch 64/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 1.1413 - acc: 0.6356 - val_loss: 1.1690 - val_acc: 0.6227\n",
            "Epoch 65/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 1.1356 - acc: 0.6366 - val_loss: 1.1634 - val_acc: 0.6232\n",
            "Epoch 66/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 1.1301 - acc: 0.6364 - val_loss: 1.1578 - val_acc: 0.6210\n",
            "Epoch 67/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 1.1246 - acc: 0.6364 - val_loss: 1.1524 - val_acc: 0.6214\n",
            "Epoch 68/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 1.1192 - acc: 0.6380 - val_loss: 1.1470 - val_acc: 0.6236\n",
            "Epoch 69/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 1.1139 - acc: 0.6393 - val_loss: 1.1416 - val_acc: 0.6236\n",
            "Epoch 70/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 1.1087 - acc: 0.6399 - val_loss: 1.1364 - val_acc: 0.6240\n",
            "Epoch 71/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 1.1036 - acc: 0.6413 - val_loss: 1.1312 - val_acc: 0.6275\n",
            "Epoch 72/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 1.0985 - acc: 0.6424 - val_loss: 1.1261 - val_acc: 0.6288\n",
            "Epoch 73/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 1.0935 - acc: 0.6421 - val_loss: 1.1211 - val_acc: 0.6284\n",
            "Epoch 74/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 1.0886 - acc: 0.6421 - val_loss: 1.1161 - val_acc: 0.6292\n",
            "Epoch 75/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 1.0837 - acc: 0.6420 - val_loss: 1.1112 - val_acc: 0.6292\n",
            "Epoch 76/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 1.0789 - acc: 0.6413 - val_loss: 1.1063 - val_acc: 0.6275\n",
            "Epoch 77/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 1.0741 - acc: 0.6417 - val_loss: 1.1015 - val_acc: 0.6449\n",
            "Epoch 78/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 1.0694 - acc: 0.6579 - val_loss: 1.0967 - val_acc: 0.6449\n",
            "Epoch 79/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 1.0647 - acc: 0.6581 - val_loss: 1.0919 - val_acc: 0.6445\n",
            "Epoch 80/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 1.0601 - acc: 0.6574 - val_loss: 1.0872 - val_acc: 0.6445\n",
            "Epoch 81/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 1.0555 - acc: 0.6573 - val_loss: 1.0826 - val_acc: 0.6445\n",
            "Epoch 82/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 1.0509 - acc: 0.6573 - val_loss: 1.0780 - val_acc: 0.6445\n",
            "Epoch 83/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 1.0465 - acc: 0.6573 - val_loss: 1.0734 - val_acc: 0.6445\n",
            "Epoch 84/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 1.0420 - acc: 0.6573 - val_loss: 1.0689 - val_acc: 0.6445\n",
            "Epoch 85/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 1.0376 - acc: 0.6573 - val_loss: 1.0643 - val_acc: 0.6445\n",
            "Epoch 86/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 1.0333 - acc: 0.6573 - val_loss: 1.0598 - val_acc: 0.6445\n",
            "Epoch 87/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 1.0289 - acc: 0.6573 - val_loss: 1.0553 - val_acc: 0.6445\n",
            "Epoch 88/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 1.0245 - acc: 0.6573 - val_loss: 1.0509 - val_acc: 0.6445\n",
            "Epoch 89/250\n",
            "184/184 [==============================] - 0s 3ms/step - loss: 1.0202 - acc: 0.6573 - val_loss: 1.0464 - val_acc: 0.6445\n",
            "Epoch 90/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 1.0159 - acc: 0.6572 - val_loss: 1.0419 - val_acc: 0.6445\n",
            "Epoch 91/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 1.0116 - acc: 0.6571 - val_loss: 1.0374 - val_acc: 0.6432\n",
            "Epoch 92/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 1.0072 - acc: 0.6568 - val_loss: 1.0329 - val_acc: 0.6432\n",
            "Epoch 93/250\n",
            "184/184 [==============================] - 0s 3ms/step - loss: 1.0029 - acc: 0.6594 - val_loss: 1.0285 - val_acc: 0.6475\n",
            "Epoch 94/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 0.9987 - acc: 0.6620 - val_loss: 1.0240 - val_acc: 0.6475\n",
            "Epoch 95/250\n",
            "184/184 [==============================] - 0s 3ms/step - loss: 0.9944 - acc: 0.6671 - val_loss: 1.0196 - val_acc: 0.6636\n",
            "Epoch 96/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 0.9901 - acc: 0.6772 - val_loss: 1.0152 - val_acc: 0.6636\n",
            "Epoch 97/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 0.9859 - acc: 0.6772 - val_loss: 1.0107 - val_acc: 0.6636\n",
            "Epoch 98/250\n",
            "184/184 [==============================] - 0s 3ms/step - loss: 0.9816 - acc: 0.6772 - val_loss: 1.0063 - val_acc: 0.6636\n",
            "Epoch 99/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 0.9774 - acc: 0.6772 - val_loss: 1.0019 - val_acc: 0.6636\n",
            "Epoch 100/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 0.9731 - acc: 0.6887 - val_loss: 0.9975 - val_acc: 0.6941\n",
            "Epoch 101/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 0.9689 - acc: 0.7032 - val_loss: 0.9931 - val_acc: 0.6941\n",
            "Epoch 102/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 0.9646 - acc: 0.7032 - val_loss: 0.9887 - val_acc: 0.6941\n",
            "Epoch 103/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 0.9604 - acc: 0.7032 - val_loss: 0.9842 - val_acc: 0.6941\n",
            "Epoch 104/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 0.9562 - acc: 0.7032 - val_loss: 0.9798 - val_acc: 0.6941\n",
            "Epoch 105/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 0.9519 - acc: 0.7032 - val_loss: 0.9754 - val_acc: 0.6941\n",
            "Epoch 106/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 0.9477 - acc: 0.7032 - val_loss: 0.9710 - val_acc: 0.6941\n",
            "Epoch 107/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 0.9435 - acc: 0.7032 - val_loss: 0.9666 - val_acc: 0.6941\n",
            "Epoch 108/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 0.9393 - acc: 0.7036 - val_loss: 0.9623 - val_acc: 0.6941\n",
            "Epoch 109/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 0.9351 - acc: 0.7051 - val_loss: 0.9579 - val_acc: 0.7023\n",
            "Epoch 110/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 0.9309 - acc: 0.7092 - val_loss: 0.9535 - val_acc: 0.7023\n",
            "Epoch 111/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 0.9267 - acc: 0.7092 - val_loss: 0.9492 - val_acc: 0.7023\n",
            "Epoch 112/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 0.9225 - acc: 0.7092 - val_loss: 0.9448 - val_acc: 0.7028\n",
            "Epoch 113/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 0.9184 - acc: 0.7095 - val_loss: 0.9406 - val_acc: 0.7028\n",
            "Epoch 114/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 0.9143 - acc: 0.7097 - val_loss: 0.9364 - val_acc: 0.7032\n",
            "Epoch 115/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 0.9103 - acc: 0.7155 - val_loss: 0.9323 - val_acc: 0.7115\n",
            "Epoch 116/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 0.9062 - acc: 0.7188 - val_loss: 0.9281 - val_acc: 0.7119\n",
            "Epoch 117/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 0.9022 - acc: 0.7188 - val_loss: 0.9240 - val_acc: 0.7119\n",
            "Epoch 118/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 0.8983 - acc: 0.7262 - val_loss: 0.9199 - val_acc: 0.7232\n",
            "Epoch 119/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 0.8943 - acc: 0.7289 - val_loss: 0.9159 - val_acc: 0.7232\n",
            "Epoch 120/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 0.8904 - acc: 0.7292 - val_loss: 0.9118 - val_acc: 0.7237\n",
            "Epoch 121/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 0.8864 - acc: 0.7298 - val_loss: 0.9078 - val_acc: 0.7237\n",
            "Epoch 122/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 0.8825 - acc: 0.7299 - val_loss: 0.9038 - val_acc: 0.7241\n",
            "Epoch 123/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 0.8787 - acc: 0.7316 - val_loss: 0.8999 - val_acc: 0.7267\n",
            "Epoch 124/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 0.8748 - acc: 0.7328 - val_loss: 0.8959 - val_acc: 0.7267\n",
            "Epoch 125/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 0.8710 - acc: 0.7359 - val_loss: 0.8920 - val_acc: 0.7367\n",
            "Epoch 126/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 0.8671 - acc: 0.7435 - val_loss: 0.8881 - val_acc: 0.7367\n",
            "Epoch 127/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 0.8633 - acc: 0.7461 - val_loss: 0.8842 - val_acc: 0.7393\n",
            "Epoch 128/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 0.8595 - acc: 0.7471 - val_loss: 0.8803 - val_acc: 0.7393\n",
            "Epoch 129/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 0.8557 - acc: 0.7471 - val_loss: 0.8765 - val_acc: 0.7393\n",
            "Epoch 130/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 0.8519 - acc: 0.7471 - val_loss: 0.8727 - val_acc: 0.7393\n",
            "Epoch 131/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 0.8482 - acc: 0.7474 - val_loss: 0.8689 - val_acc: 0.7411\n",
            "Epoch 132/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 0.8445 - acc: 0.7483 - val_loss: 0.8651 - val_acc: 0.7411\n",
            "Epoch 133/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 0.8407 - acc: 0.7483 - val_loss: 0.8614 - val_acc: 0.7411\n",
            "Epoch 134/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 0.8370 - acc: 0.7488 - val_loss: 0.8576 - val_acc: 0.7419\n",
            "Epoch 135/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 0.8334 - acc: 0.7520 - val_loss: 0.8539 - val_acc: 0.7419\n",
            "Epoch 136/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 0.8297 - acc: 0.7522 - val_loss: 0.8503 - val_acc: 0.7419\n",
            "Epoch 137/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 0.8261 - acc: 0.7522 - val_loss: 0.8466 - val_acc: 0.7419\n",
            "Epoch 138/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 0.8225 - acc: 0.7532 - val_loss: 0.8430 - val_acc: 0.7428\n",
            "Epoch 139/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 0.8189 - acc: 0.7544 - val_loss: 0.8394 - val_acc: 0.7428\n",
            "Epoch 140/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 0.8153 - acc: 0.7547 - val_loss: 0.8358 - val_acc: 0.7428\n",
            "Epoch 141/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 0.8118 - acc: 0.7559 - val_loss: 0.8322 - val_acc: 0.7437\n",
            "Epoch 142/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 0.8082 - acc: 0.7562 - val_loss: 0.8287 - val_acc: 0.7437\n",
            "Epoch 143/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 0.8047 - acc: 0.7562 - val_loss: 0.8252 - val_acc: 0.7428\n",
            "Epoch 144/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 0.8012 - acc: 0.7565 - val_loss: 0.8217 - val_acc: 0.7472\n",
            "Epoch 145/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 0.7977 - acc: 0.7655 - val_loss: 0.8182 - val_acc: 0.7533\n",
            "Epoch 146/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 0.7942 - acc: 0.7700 - val_loss: 0.8147 - val_acc: 0.7615\n",
            "Epoch 147/250\n",
            "184/184 [==============================] - 0s 3ms/step - loss: 0.7908 - acc: 0.7721 - val_loss: 0.8113 - val_acc: 0.7615\n",
            "Epoch 148/250\n",
            "184/184 [==============================] - 1s 3ms/step - loss: 0.7873 - acc: 0.7729 - val_loss: 0.8078 - val_acc: 0.7624\n",
            "Epoch 149/250\n",
            "184/184 [==============================] - 1s 3ms/step - loss: 0.7839 - acc: 0.7741 - val_loss: 0.8044 - val_acc: 0.7624\n",
            "Epoch 150/250\n",
            "184/184 [==============================] - 1s 3ms/step - loss: 0.7805 - acc: 0.7741 - val_loss: 0.8010 - val_acc: 0.7624\n",
            "Epoch 151/250\n",
            "184/184 [==============================] - 1s 3ms/step - loss: 0.7771 - acc: 0.7768 - val_loss: 0.7976 - val_acc: 0.7681\n",
            "Epoch 152/250\n",
            "184/184 [==============================] - 1s 3ms/step - loss: 0.7737 - acc: 0.7778 - val_loss: 0.7942 - val_acc: 0.7702\n",
            "Epoch 153/250\n",
            "184/184 [==============================] - 1s 3ms/step - loss: 0.7703 - acc: 0.7797 - val_loss: 0.7909 - val_acc: 0.7702\n",
            "Epoch 154/250\n",
            "184/184 [==============================] - 1s 3ms/step - loss: 0.7669 - acc: 0.7817 - val_loss: 0.7875 - val_acc: 0.7724\n",
            "Epoch 155/250\n",
            "184/184 [==============================] - 1s 3ms/step - loss: 0.7635 - acc: 0.7840 - val_loss: 0.7842 - val_acc: 0.7750\n",
            "Epoch 156/250\n",
            "184/184 [==============================] - 1s 3ms/step - loss: 0.7602 - acc: 0.7848 - val_loss: 0.7809 - val_acc: 0.7750\n",
            "Epoch 157/250\n",
            "184/184 [==============================] - 1s 3ms/step - loss: 0.7568 - acc: 0.7860 - val_loss: 0.7775 - val_acc: 0.7763\n",
            "Epoch 158/250\n",
            "184/184 [==============================] - 1s 3ms/step - loss: 0.7534 - acc: 0.7867 - val_loss: 0.7742 - val_acc: 0.7763\n",
            "Epoch 159/250\n",
            "184/184 [==============================] - 1s 3ms/step - loss: 0.7500 - acc: 0.7867 - val_loss: 0.7708 - val_acc: 0.7763\n",
            "Epoch 160/250\n",
            "184/184 [==============================] - 1s 3ms/step - loss: 0.7467 - acc: 0.7868 - val_loss: 0.7675 - val_acc: 0.7768\n",
            "Epoch 161/250\n",
            "184/184 [==============================] - 1s 3ms/step - loss: 0.7433 - acc: 0.7871 - val_loss: 0.7642 - val_acc: 0.7776\n",
            "Epoch 162/250\n",
            "184/184 [==============================] - 1s 3ms/step - loss: 0.7400 - acc: 0.7896 - val_loss: 0.7609 - val_acc: 0.7776\n",
            "Epoch 163/250\n",
            "184/184 [==============================] - 1s 3ms/step - loss: 0.7366 - acc: 0.7897 - val_loss: 0.7576 - val_acc: 0.7776\n",
            "Epoch 164/250\n",
            "184/184 [==============================] - 1s 3ms/step - loss: 0.7333 - acc: 0.7897 - val_loss: 0.7543 - val_acc: 0.7776\n",
            "Epoch 165/250\n",
            "184/184 [==============================] - 0s 3ms/step - loss: 0.7300 - acc: 0.7898 - val_loss: 0.7510 - val_acc: 0.7776\n",
            "Epoch 166/250\n",
            "184/184 [==============================] - 1s 3ms/step - loss: 0.7267 - acc: 0.7906 - val_loss: 0.7477 - val_acc: 0.7811\n",
            "Epoch 167/250\n",
            "184/184 [==============================] - 1s 3ms/step - loss: 0.7234 - acc: 0.7923 - val_loss: 0.7445 - val_acc: 0.7833\n",
            "Epoch 168/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 0.7200 - acc: 0.7945 - val_loss: 0.7412 - val_acc: 0.7833\n",
            "Epoch 169/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 0.7167 - acc: 0.7949 - val_loss: 0.7379 - val_acc: 0.7833\n",
            "Epoch 170/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 0.7134 - acc: 0.7951 - val_loss: 0.7347 - val_acc: 0.7837\n",
            "Epoch 171/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 0.7101 - acc: 0.7964 - val_loss: 0.7314 - val_acc: 0.7837\n",
            "Epoch 172/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 0.7069 - acc: 0.7966 - val_loss: 0.7281 - val_acc: 0.7837\n",
            "Epoch 173/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 0.7036 - acc: 0.7975 - val_loss: 0.7249 - val_acc: 0.7846\n",
            "Epoch 174/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 0.7003 - acc: 0.7979 - val_loss: 0.7216 - val_acc: 0.7846\n",
            "Epoch 175/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 0.6970 - acc: 0.8008 - val_loss: 0.7184 - val_acc: 0.7950\n",
            "Epoch 176/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 0.6937 - acc: 0.8065 - val_loss: 0.7151 - val_acc: 0.7950\n",
            "Epoch 177/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 0.6904 - acc: 0.8065 - val_loss: 0.7118 - val_acc: 0.7950\n",
            "Epoch 178/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 0.6871 - acc: 0.8070 - val_loss: 0.7085 - val_acc: 0.7972\n",
            "Epoch 179/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 0.6837 - acc: 0.8078 - val_loss: 0.7052 - val_acc: 0.7972\n",
            "Epoch 180/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 0.6804 - acc: 0.8078 - val_loss: 0.7019 - val_acc: 0.7981\n",
            "Epoch 181/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 0.6771 - acc: 0.8082 - val_loss: 0.6986 - val_acc: 0.7998\n",
            "Epoch 182/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 0.6738 - acc: 0.8089 - val_loss: 0.6953 - val_acc: 0.7998\n",
            "Epoch 183/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 0.6704 - acc: 0.8089 - val_loss: 0.6920 - val_acc: 0.7998\n",
            "Epoch 184/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 0.6671 - acc: 0.8089 - val_loss: 0.6888 - val_acc: 0.7998\n",
            "Epoch 185/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 0.6638 - acc: 0.8089 - val_loss: 0.6855 - val_acc: 0.7998\n",
            "Epoch 186/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 0.6605 - acc: 0.8089 - val_loss: 0.6822 - val_acc: 0.7998\n",
            "Epoch 187/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 0.6572 - acc: 0.8089 - val_loss: 0.6789 - val_acc: 0.7998\n",
            "Epoch 188/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 0.6539 - acc: 0.8089 - val_loss: 0.6756 - val_acc: 0.7998\n",
            "Epoch 189/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 0.6506 - acc: 0.8109 - val_loss: 0.6724 - val_acc: 0.8033\n",
            "Epoch 190/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 0.6473 - acc: 0.8140 - val_loss: 0.6691 - val_acc: 0.8037\n",
            "Epoch 191/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 0.6440 - acc: 0.8146 - val_loss: 0.6658 - val_acc: 0.8037\n",
            "Epoch 192/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 0.6407 - acc: 0.8146 - val_loss: 0.6626 - val_acc: 0.8037\n",
            "Epoch 193/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 0.6374 - acc: 0.8146 - val_loss: 0.6593 - val_acc: 0.8037\n",
            "Epoch 194/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 0.6341 - acc: 0.8148 - val_loss: 0.6561 - val_acc: 0.8037\n",
            "Epoch 195/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 0.6309 - acc: 0.8150 - val_loss: 0.6528 - val_acc: 0.8037\n",
            "Epoch 196/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 0.6276 - acc: 0.8151 - val_loss: 0.6496 - val_acc: 0.8037\n",
            "Epoch 197/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 0.6243 - acc: 0.8160 - val_loss: 0.6463 - val_acc: 0.8103\n",
            "Epoch 198/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 0.6211 - acc: 0.8240 - val_loss: 0.6431 - val_acc: 0.8111\n",
            "Epoch 199/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 0.6178 - acc: 0.8254 - val_loss: 0.6399 - val_acc: 0.8124\n",
            "Epoch 200/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 0.6146 - acc: 0.8340 - val_loss: 0.6367 - val_acc: 0.8207\n",
            "Epoch 201/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 0.6113 - acc: 0.8369 - val_loss: 0.6334 - val_acc: 0.8207\n",
            "Epoch 202/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 0.6081 - acc: 0.8373 - val_loss: 0.6302 - val_acc: 0.8225\n",
            "Epoch 203/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 0.6048 - acc: 0.8380 - val_loss: 0.6270 - val_acc: 0.8225\n",
            "Epoch 204/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 0.6016 - acc: 0.8399 - val_loss: 0.6238 - val_acc: 0.8251\n",
            "Epoch 205/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 0.5984 - acc: 0.8420 - val_loss: 0.6206 - val_acc: 0.8255\n",
            "Epoch 206/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 0.5951 - acc: 0.8424 - val_loss: 0.6174 - val_acc: 0.8259\n",
            "Epoch 207/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 0.5919 - acc: 0.8431 - val_loss: 0.6143 - val_acc: 0.8290\n",
            "Epoch 208/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 0.5887 - acc: 0.8449 - val_loss: 0.6111 - val_acc: 0.8303\n",
            "Epoch 209/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 0.5855 - acc: 0.8452 - val_loss: 0.6079 - val_acc: 0.8303\n",
            "Epoch 210/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 0.5824 - acc: 0.8492 - val_loss: 0.6048 - val_acc: 0.8399\n",
            "Epoch 211/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 0.5792 - acc: 0.8539 - val_loss: 0.6016 - val_acc: 0.8403\n",
            "Epoch 212/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 0.5760 - acc: 0.8538 - val_loss: 0.5985 - val_acc: 0.8403\n",
            "Epoch 213/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 0.5728 - acc: 0.8540 - val_loss: 0.5954 - val_acc: 0.8403\n",
            "Epoch 214/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 0.5697 - acc: 0.8544 - val_loss: 0.5922 - val_acc: 0.8425\n",
            "Epoch 215/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 0.5666 - acc: 0.8549 - val_loss: 0.5891 - val_acc: 0.8425\n",
            "Epoch 216/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 0.5634 - acc: 0.8550 - val_loss: 0.5860 - val_acc: 0.8433\n",
            "Epoch 217/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 0.5603 - acc: 0.8551 - val_loss: 0.5829 - val_acc: 0.8433\n",
            "Epoch 218/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 0.5572 - acc: 0.8551 - val_loss: 0.5799 - val_acc: 0.8433\n",
            "Epoch 219/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 0.5541 - acc: 0.8563 - val_loss: 0.5768 - val_acc: 0.8438\n",
            "Epoch 220/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 0.5510 - acc: 0.8567 - val_loss: 0.5737 - val_acc: 0.8464\n",
            "Epoch 221/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 0.5479 - acc: 0.8573 - val_loss: 0.5707 - val_acc: 0.8464\n",
            "Epoch 222/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 0.5449 - acc: 0.8572 - val_loss: 0.5676 - val_acc: 0.8464\n",
            "Epoch 223/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 0.5418 - acc: 0.8572 - val_loss: 0.5646 - val_acc: 0.8464\n",
            "Epoch 224/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 0.5388 - acc: 0.8573 - val_loss: 0.5616 - val_acc: 0.8464\n",
            "Epoch 225/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 0.5357 - acc: 0.8573 - val_loss: 0.5586 - val_acc: 0.8464\n",
            "Epoch 226/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 0.5327 - acc: 0.8573 - val_loss: 0.5556 - val_acc: 0.8464\n",
            "Epoch 227/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 0.5297 - acc: 0.8577 - val_loss: 0.5526 - val_acc: 0.8468\n",
            "Epoch 228/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 0.5267 - acc: 0.8585 - val_loss: 0.5496 - val_acc: 0.8468\n",
            "Epoch 229/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 0.5237 - acc: 0.8586 - val_loss: 0.5466 - val_acc: 0.8486\n",
            "Epoch 230/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 0.5207 - acc: 0.8595 - val_loss: 0.5437 - val_acc: 0.8494\n",
            "Epoch 231/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 0.5178 - acc: 0.8608 - val_loss: 0.5408 - val_acc: 0.8494\n",
            "Epoch 232/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 0.5148 - acc: 0.8608 - val_loss: 0.5378 - val_acc: 0.8494\n",
            "Epoch 233/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 0.5119 - acc: 0.8608 - val_loss: 0.5349 - val_acc: 0.8494\n",
            "Epoch 234/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 0.5090 - acc: 0.8608 - val_loss: 0.5320 - val_acc: 0.8494\n",
            "Epoch 235/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 0.5061 - acc: 0.8612 - val_loss: 0.5291 - val_acc: 0.8542\n",
            "Epoch 236/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 0.5032 - acc: 0.8647 - val_loss: 0.5262 - val_acc: 0.8560\n",
            "Epoch 237/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 0.5003 - acc: 0.8674 - val_loss: 0.5234 - val_acc: 0.8621\n",
            "Epoch 238/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 0.4974 - acc: 0.8702 - val_loss: 0.5205 - val_acc: 0.8621\n",
            "Epoch 239/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 0.4946 - acc: 0.8717 - val_loss: 0.5177 - val_acc: 0.8642\n",
            "Epoch 240/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 0.4917 - acc: 0.8718 - val_loss: 0.5148 - val_acc: 0.8642\n",
            "Epoch 241/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 0.4889 - acc: 0.8718 - val_loss: 0.5120 - val_acc: 0.8642\n",
            "Epoch 242/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 0.4861 - acc: 0.8718 - val_loss: 0.5092 - val_acc: 0.8642\n",
            "Epoch 243/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 0.4833 - acc: 0.8741 - val_loss: 0.5064 - val_acc: 0.8677\n",
            "Epoch 244/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 0.4805 - acc: 0.8749 - val_loss: 0.5037 - val_acc: 0.8681\n",
            "Epoch 245/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 0.4777 - acc: 0.8774 - val_loss: 0.5009 - val_acc: 0.8725\n",
            "Epoch 246/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 0.4750 - acc: 0.8790 - val_loss: 0.4982 - val_acc: 0.8725\n",
            "Epoch 247/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 0.4722 - acc: 0.8797 - val_loss: 0.4954 - val_acc: 0.8760\n",
            "Epoch 248/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 0.4695 - acc: 0.8833 - val_loss: 0.4927 - val_acc: 0.8760\n",
            "Epoch 249/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 0.4668 - acc: 0.8838 - val_loss: 0.4900 - val_acc: 0.8760\n",
            "Epoch 250/250\n",
            "184/184 [==============================] - 0s 2ms/step - loss: 0.4641 - acc: 0.8838 - val_loss: 0.4873 - val_acc: 0.8760\n",
            "<bound method Model.summary of <tensorflow.python.keras.engine.sequential.Sequential object at 0x7fd83d57f4e0>>\n",
            "Test Accuracy:\n",
            "2068/2068 [==============================] - 3s 2ms/step - loss: 0.4640 - acc: 0.8811\n",
            "Random Sample Accuracy:\n",
            "30/30 [==============================] - 0s 2ms/step - loss: 0.9007 - acc: 0.7013\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 1080x432 with 2 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "H843Nzhp68kH"
      },
      "source": [
        "model5_params = {'activation 1': 'relu', \n",
        "                 'activation 2': 'relu', \n",
        "                 'activation 3': 'softmax', \n",
        "                 'opt': tf.keras.optimizers.SGD(learning_rate=0.001),\n",
        "                 'loss': 'categorical_crossentropy',\n",
        "                 'epochs': 270, \n",
        "                 'batch_size': 40, \n",
        "                 }"
      ],
      "execution_count": 48,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 1000
        },
        "id": "zx-thXl37FTR",
        "outputId": "cf79a66b-a6db-4ba3-bef3-bb1f307110fe"
      },
      "source": [
        "model5_history, model5 = create_model(train, label_train, test, label_test, model5_params)"
      ],
      "execution_count": 49,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Epoch 1/270\n",
            "230/230 [==============================] - 1s 3ms/step - loss: 2.5883 - acc: 0.3399 - val_loss: 2.5324 - val_acc: 0.5096\n",
            "Epoch 2/270\n",
            "230/230 [==============================] - 0s 2ms/step - loss: 2.4650 - acc: 0.5212 - val_loss: 2.4024 - val_acc: 0.5109\n",
            "Epoch 3/270\n",
            "230/230 [==============================] - 0s 2ms/step - loss: 2.3273 - acc: 0.5215 - val_loss: 2.2667 - val_acc: 0.5109\n",
            "Epoch 4/270\n",
            "230/230 [==============================] - 1s 2ms/step - loss: 2.1859 - acc: 0.5215 - val_loss: 2.1280 - val_acc: 0.5109\n",
            "Epoch 5/270\n",
            "230/230 [==============================] - 1s 2ms/step - loss: 2.0423 - acc: 0.5215 - val_loss: 1.9900 - val_acc: 0.5109\n",
            "Epoch 6/270\n",
            "230/230 [==============================] - 0s 2ms/step - loss: 1.9058 - acc: 0.5215 - val_loss: 1.8670 - val_acc: 0.5109\n",
            "Epoch 7/270\n",
            "230/230 [==============================] - 1s 2ms/step - loss: 1.7925 - acc: 0.5215 - val_loss: 1.7736 - val_acc: 0.5109\n",
            "Epoch 8/270\n",
            "230/230 [==============================] - 0s 2ms/step - loss: 1.7119 - acc: 0.5215 - val_loss: 1.7115 - val_acc: 0.5109\n",
            "Epoch 9/270\n",
            "230/230 [==============================] - 1s 2ms/step - loss: 1.6597 - acc: 0.5215 - val_loss: 1.6718 - val_acc: 0.5109\n",
            "Epoch 10/270\n",
            "230/230 [==============================] - 1s 2ms/step - loss: 1.6253 - acc: 0.5215 - val_loss: 1.6442 - val_acc: 0.5109\n",
            "Epoch 11/270\n",
            "230/230 [==============================] - 0s 2ms/step - loss: 1.6001 - acc: 0.5215 - val_loss: 1.6225 - val_acc: 0.5109\n",
            "Epoch 12/270\n",
            "230/230 [==============================] - 0s 2ms/step - loss: 1.5795 - acc: 0.5215 - val_loss: 1.6037 - val_acc: 0.5109\n",
            "Epoch 13/270\n",
            "230/230 [==============================] - 1s 2ms/step - loss: 1.5612 - acc: 0.5215 - val_loss: 1.5864 - val_acc: 0.5109\n",
            "Epoch 14/270\n",
            "230/230 [==============================] - 0s 2ms/step - loss: 1.5443 - acc: 0.5215 - val_loss: 1.5702 - val_acc: 0.5109\n",
            "Epoch 15/270\n",
            "230/230 [==============================] - 1s 2ms/step - loss: 1.5282 - acc: 0.5215 - val_loss: 1.5544 - val_acc: 0.5109\n",
            "Epoch 16/270\n",
            "230/230 [==============================] - 1s 2ms/step - loss: 1.5126 - acc: 0.5215 - val_loss: 1.5389 - val_acc: 0.5109\n",
            "Epoch 17/270\n",
            "230/230 [==============================] - 1s 2ms/step - loss: 1.4973 - acc: 0.5215 - val_loss: 1.5239 - val_acc: 0.5109\n",
            "Epoch 18/270\n",
            "230/230 [==============================] - 0s 2ms/step - loss: 1.4826 - acc: 0.5215 - val_loss: 1.5094 - val_acc: 0.5109\n",
            "Epoch 19/270\n",
            "230/230 [==============================] - 0s 2ms/step - loss: 1.4685 - acc: 0.5215 - val_loss: 1.4954 - val_acc: 0.5109\n",
            "Epoch 20/270\n",
            "230/230 [==============================] - 0s 2ms/step - loss: 1.4547 - acc: 0.5215 - val_loss: 1.4818 - val_acc: 0.5109\n",
            "Epoch 21/270\n",
            "230/230 [==============================] - 1s 2ms/step - loss: 1.4413 - acc: 0.5215 - val_loss: 1.4683 - val_acc: 0.5109\n",
            "Epoch 22/270\n",
            "230/230 [==============================] - 0s 2ms/step - loss: 1.4280 - acc: 0.5215 - val_loss: 1.4548 - val_acc: 0.5109\n",
            "Epoch 23/270\n",
            "230/230 [==============================] - 1s 2ms/step - loss: 1.4147 - acc: 0.5215 - val_loss: 1.4414 - val_acc: 0.5109\n",
            "Epoch 24/270\n",
            "230/230 [==============================] - 0s 2ms/step - loss: 1.4016 - acc: 0.5215 - val_loss: 1.4283 - val_acc: 0.5109\n",
            "Epoch 25/270\n",
            "230/230 [==============================] - 1s 2ms/step - loss: 1.3888 - acc: 0.5215 - val_loss: 1.4154 - val_acc: 0.5109\n",
            "Epoch 26/270\n",
            "230/230 [==============================] - 0s 2ms/step - loss: 1.3762 - acc: 0.5215 - val_loss: 1.4027 - val_acc: 0.5109\n",
            "Epoch 27/270\n",
            "230/230 [==============================] - 1s 2ms/step - loss: 1.3640 - acc: 0.5215 - val_loss: 1.3905 - val_acc: 0.5109\n",
            "Epoch 28/270\n",
            "230/230 [==============================] - 0s 2ms/step - loss: 1.3521 - acc: 0.5215 - val_loss: 1.3786 - val_acc: 0.5109\n",
            "Epoch 29/270\n",
            "230/230 [==============================] - 0s 2ms/step - loss: 1.3406 - acc: 0.5215 - val_loss: 1.3672 - val_acc: 0.5109\n",
            "Epoch 30/270\n",
            "230/230 [==============================] - 0s 2ms/step - loss: 1.3296 - acc: 0.5215 - val_loss: 1.3562 - val_acc: 0.5109\n",
            "Epoch 31/270\n",
            "230/230 [==============================] - 0s 2ms/step - loss: 1.3187 - acc: 0.5215 - val_loss: 1.3453 - val_acc: 0.5109\n",
            "Epoch 32/270\n",
            "230/230 [==============================] - 1s 2ms/step - loss: 1.3081 - acc: 0.5215 - val_loss: 1.3345 - val_acc: 0.5109\n",
            "Epoch 33/270\n",
            "230/230 [==============================] - 0s 2ms/step - loss: 1.2976 - acc: 0.5215 - val_loss: 1.3241 - val_acc: 0.5109\n",
            "Epoch 34/270\n",
            "230/230 [==============================] - 1s 2ms/step - loss: 1.2874 - acc: 0.5215 - val_loss: 1.3138 - val_acc: 0.5109\n",
            "Epoch 35/270\n",
            "230/230 [==============================] - 1s 2ms/step - loss: 1.2773 - acc: 0.5215 - val_loss: 1.3035 - val_acc: 0.5109\n",
            "Epoch 36/270\n",
            "230/230 [==============================] - 1s 2ms/step - loss: 1.2672 - acc: 0.5215 - val_loss: 1.2934 - val_acc: 0.5109\n",
            "Epoch 37/270\n",
            "230/230 [==============================] - 0s 2ms/step - loss: 1.2574 - acc: 0.5215 - val_loss: 1.2835 - val_acc: 0.5109\n",
            "Epoch 38/270\n",
            "230/230 [==============================] - 0s 2ms/step - loss: 1.2477 - acc: 0.5364 - val_loss: 1.2737 - val_acc: 0.5335\n",
            "Epoch 39/270\n",
            "230/230 [==============================] - 0s 2ms/step - loss: 1.2381 - acc: 0.5409 - val_loss: 1.2641 - val_acc: 0.5335\n",
            "Epoch 40/270\n",
            "230/230 [==============================] - 1s 2ms/step - loss: 1.2287 - acc: 0.5409 - val_loss: 1.2547 - val_acc: 0.5335\n",
            "Epoch 41/270\n",
            "230/230 [==============================] - 1s 2ms/step - loss: 1.2195 - acc: 0.5414 - val_loss: 1.2453 - val_acc: 0.5387\n",
            "Epoch 42/270\n",
            "230/230 [==============================] - 1s 2ms/step - loss: 1.2103 - acc: 0.5455 - val_loss: 1.2362 - val_acc: 0.5387\n",
            "Epoch 43/270\n",
            "230/230 [==============================] - 0s 2ms/step - loss: 1.2014 - acc: 0.5455 - val_loss: 1.2271 - val_acc: 0.5387\n",
            "Epoch 44/270\n",
            "230/230 [==============================] - 0s 2ms/step - loss: 1.1926 - acc: 0.5455 - val_loss: 1.2182 - val_acc: 0.5387\n",
            "Epoch 45/270\n",
            "230/230 [==============================] - 0s 2ms/step - loss: 1.1838 - acc: 0.5540 - val_loss: 1.2094 - val_acc: 0.5561\n",
            "Epoch 46/270\n",
            "230/230 [==============================] - 0s 2ms/step - loss: 1.1752 - acc: 0.5759 - val_loss: 1.2008 - val_acc: 0.5679\n",
            "Epoch 47/270\n",
            "230/230 [==============================] - 1s 2ms/step - loss: 1.1667 - acc: 0.5779 - val_loss: 1.1921 - val_acc: 0.5679\n",
            "Epoch 48/270\n",
            "230/230 [==============================] - 0s 2ms/step - loss: 1.1583 - acc: 0.5822 - val_loss: 1.1837 - val_acc: 0.5792\n",
            "Epoch 49/270\n",
            "230/230 [==============================] - 1s 2ms/step - loss: 1.1500 - acc: 0.5976 - val_loss: 1.1752 - val_acc: 0.5836\n",
            "Epoch 50/270\n",
            "230/230 [==============================] - 0s 2ms/step - loss: 1.1418 - acc: 0.6020 - val_loss: 1.1670 - val_acc: 0.5896\n",
            "Epoch 51/270\n",
            "230/230 [==============================] - 1s 2ms/step - loss: 1.1337 - acc: 0.6035 - val_loss: 1.1588 - val_acc: 0.5901\n",
            "Epoch 52/270\n",
            "230/230 [==============================] - 0s 2ms/step - loss: 1.1258 - acc: 0.6220 - val_loss: 1.1507 - val_acc: 0.6205\n",
            "Epoch 53/270\n",
            "230/230 [==============================] - 0s 2ms/step - loss: 1.1179 - acc: 0.6330 - val_loss: 1.1427 - val_acc: 0.6240\n",
            "Epoch 54/270\n",
            "230/230 [==============================] - 1s 2ms/step - loss: 1.1100 - acc: 0.6400 - val_loss: 1.1347 - val_acc: 0.6310\n",
            "Epoch 55/270\n",
            "230/230 [==============================] - 1s 2ms/step - loss: 1.1023 - acc: 0.6415 - val_loss: 1.1268 - val_acc: 0.6305\n",
            "Epoch 56/270\n",
            "230/230 [==============================] - 1s 2ms/step - loss: 1.0946 - acc: 0.6483 - val_loss: 1.1191 - val_acc: 0.6462\n",
            "Epoch 57/270\n",
            "230/230 [==============================] - 1s 2ms/step - loss: 1.0871 - acc: 0.6627 - val_loss: 1.1115 - val_acc: 0.6628\n",
            "Epoch 58/270\n",
            "230/230 [==============================] - 1s 2ms/step - loss: 1.0797 - acc: 0.6731 - val_loss: 1.1039 - val_acc: 0.6636\n",
            "Epoch 59/270\n",
            "230/230 [==============================] - 1s 2ms/step - loss: 1.0723 - acc: 0.6869 - val_loss: 1.0964 - val_acc: 0.6928\n",
            "Epoch 60/270\n",
            "230/230 [==============================] - 1s 2ms/step - loss: 1.0651 - acc: 0.6994 - val_loss: 1.0890 - val_acc: 0.6928\n",
            "Epoch 61/270\n",
            "230/230 [==============================] - 1s 2ms/step - loss: 1.0578 - acc: 0.7003 - val_loss: 1.0816 - val_acc: 0.6963\n",
            "Epoch 62/270\n",
            "230/230 [==============================] - 1s 2ms/step - loss: 1.0506 - acc: 0.7024 - val_loss: 1.0743 - val_acc: 0.6989\n",
            "Epoch 63/270\n",
            "230/230 [==============================] - 1s 2ms/step - loss: 1.0435 - acc: 0.7030 - val_loss: 1.0670 - val_acc: 0.6989\n",
            "Epoch 64/270\n",
            "230/230 [==============================] - 1s 2ms/step - loss: 1.0365 - acc: 0.7025 - val_loss: 1.0597 - val_acc: 0.7010\n",
            "Epoch 65/270\n",
            "230/230 [==============================] - 1s 2ms/step - loss: 1.0295 - acc: 0.7060 - val_loss: 1.0526 - val_acc: 0.7023\n",
            "Epoch 66/270\n",
            "230/230 [==============================] - 1s 2ms/step - loss: 1.0225 - acc: 0.7078 - val_loss: 1.0455 - val_acc: 0.7041\n",
            "Epoch 67/270\n",
            "230/230 [==============================] - 1s 2ms/step - loss: 1.0156 - acc: 0.7091 - val_loss: 1.0385 - val_acc: 0.7041\n",
            "Epoch 68/270\n",
            "230/230 [==============================] - 1s 2ms/step - loss: 1.0088 - acc: 0.7089 - val_loss: 1.0316 - val_acc: 0.7010\n",
            "Epoch 69/270\n",
            "230/230 [==============================] - 1s 2ms/step - loss: 1.0021 - acc: 0.7084 - val_loss: 1.0246 - val_acc: 0.7010\n",
            "Epoch 70/270\n",
            "230/230 [==============================] - 1s 2ms/step - loss: 0.9953 - acc: 0.7084 - val_loss: 1.0177 - val_acc: 0.7010\n",
            "Epoch 71/270\n",
            "230/230 [==============================] - 1s 2ms/step - loss: 0.9885 - acc: 0.7091 - val_loss: 1.0107 - val_acc: 0.7019\n",
            "Epoch 72/270\n",
            "230/230 [==============================] - 1s 2ms/step - loss: 0.9817 - acc: 0.7096 - val_loss: 1.0038 - val_acc: 0.7019\n",
            "Epoch 73/270\n",
            "230/230 [==============================] - 1s 2ms/step - loss: 0.9750 - acc: 0.7085 - val_loss: 0.9969 - val_acc: 0.7019\n",
            "Epoch 74/270\n",
            "230/230 [==============================] - 0s 2ms/step - loss: 0.9683 - acc: 0.7085 - val_loss: 0.9901 - val_acc: 0.7019\n",
            "Epoch 75/270\n",
            "230/230 [==============================] - 1s 2ms/step - loss: 0.9616 - acc: 0.7080 - val_loss: 0.9833 - val_acc: 0.7019\n",
            "Epoch 76/270\n",
            "230/230 [==============================] - 1s 2ms/step - loss: 0.9549 - acc: 0.7080 - val_loss: 0.9765 - val_acc: 0.7019\n",
            "Epoch 77/270\n",
            "230/230 [==============================] - 1s 2ms/step - loss: 0.9483 - acc: 0.7080 - val_loss: 0.9697 - val_acc: 0.7019\n",
            "Epoch 78/270\n",
            "230/230 [==============================] - 0s 2ms/step - loss: 0.9417 - acc: 0.7091 - val_loss: 0.9630 - val_acc: 0.7037\n",
            "Epoch 79/270\n",
            "230/230 [==============================] - 1s 2ms/step - loss: 0.9352 - acc: 0.7093 - val_loss: 0.9563 - val_acc: 0.7037\n",
            "Epoch 80/270\n",
            "230/230 [==============================] - 1s 2ms/step - loss: 0.9287 - acc: 0.7097 - val_loss: 0.9497 - val_acc: 0.7058\n",
            "Epoch 81/270\n",
            "230/230 [==============================] - 0s 2ms/step - loss: 0.9223 - acc: 0.7119 - val_loss: 0.9432 - val_acc: 0.7058\n",
            "Epoch 82/270\n",
            "230/230 [==============================] - 1s 2ms/step - loss: 0.9159 - acc: 0.7119 - val_loss: 0.9367 - val_acc: 0.7058\n",
            "Epoch 83/270\n",
            "230/230 [==============================] - 1s 2ms/step - loss: 0.9096 - acc: 0.7119 - val_loss: 0.9303 - val_acc: 0.7058\n",
            "Epoch 84/270\n",
            "230/230 [==============================] - 1s 2ms/step - loss: 0.9033 - acc: 0.7119 - val_loss: 0.9239 - val_acc: 0.7058\n",
            "Epoch 85/270\n",
            "230/230 [==============================] - 1s 2ms/step - loss: 0.8971 - acc: 0.7130 - val_loss: 0.9175 - val_acc: 0.7115\n",
            "Epoch 86/270\n",
            "230/230 [==============================] - 0s 2ms/step - loss: 0.8909 - acc: 0.7176 - val_loss: 0.9113 - val_acc: 0.7124\n",
            "Epoch 87/270\n",
            "230/230 [==============================] - 0s 2ms/step - loss: 0.8848 - acc: 0.7183 - val_loss: 0.9051 - val_acc: 0.7124\n",
            "Epoch 88/270\n",
            "230/230 [==============================] - 0s 2ms/step - loss: 0.8787 - acc: 0.7183 - val_loss: 0.8989 - val_acc: 0.7124\n",
            "Epoch 89/270\n",
            "230/230 [==============================] - 0s 2ms/step - loss: 0.8728 - acc: 0.7183 - val_loss: 0.8929 - val_acc: 0.7124\n",
            "Epoch 90/270\n",
            "230/230 [==============================] - 0s 2ms/step - loss: 0.8668 - acc: 0.7183 - val_loss: 0.8870 - val_acc: 0.7124\n",
            "Epoch 91/270\n",
            "230/230 [==============================] - 0s 2ms/step - loss: 0.8610 - acc: 0.7225 - val_loss: 0.8811 - val_acc: 0.7167\n",
            "Epoch 92/270\n",
            "230/230 [==============================] - 1s 2ms/step - loss: 0.8552 - acc: 0.7254 - val_loss: 0.8752 - val_acc: 0.7228\n",
            "Epoch 93/270\n",
            "230/230 [==============================] - 0s 2ms/step - loss: 0.8494 - acc: 0.7270 - val_loss: 0.8694 - val_acc: 0.7232\n",
            "Epoch 94/270\n",
            "230/230 [==============================] - 0s 2ms/step - loss: 0.8436 - acc: 0.7273 - val_loss: 0.8636 - val_acc: 0.7232\n",
            "Epoch 95/270\n",
            "230/230 [==============================] - 1s 2ms/step - loss: 0.8379 - acc: 0.7279 - val_loss: 0.8579 - val_acc: 0.7263\n",
            "Epoch 96/270\n",
            "230/230 [==============================] - 0s 2ms/step - loss: 0.8322 - acc: 0.7301 - val_loss: 0.8522 - val_acc: 0.7263\n",
            "Epoch 97/270\n",
            "230/230 [==============================] - 1s 2ms/step - loss: 0.8266 - acc: 0.7369 - val_loss: 0.8465 - val_acc: 0.7341\n",
            "Epoch 98/270\n",
            "230/230 [==============================] - 0s 2ms/step - loss: 0.8210 - acc: 0.7421 - val_loss: 0.8409 - val_acc: 0.7424\n",
            "Epoch 99/270\n",
            "230/230 [==============================] - 0s 2ms/step - loss: 0.8155 - acc: 0.7486 - val_loss: 0.8354 - val_acc: 0.7424\n",
            "Epoch 100/270\n",
            "230/230 [==============================] - 1s 2ms/step - loss: 0.8099 - acc: 0.7489 - val_loss: 0.8298 - val_acc: 0.7424\n",
            "Epoch 101/270\n",
            "230/230 [==============================] - 0s 2ms/step - loss: 0.8045 - acc: 0.7491 - val_loss: 0.8243 - val_acc: 0.7424\n",
            "Epoch 102/270\n",
            "230/230 [==============================] - 1s 2ms/step - loss: 0.7990 - acc: 0.7491 - val_loss: 0.8189 - val_acc: 0.7424\n",
            "Epoch 103/270\n",
            "230/230 [==============================] - 1s 2ms/step - loss: 0.7936 - acc: 0.7526 - val_loss: 0.8135 - val_acc: 0.7472\n",
            "Epoch 104/270\n",
            "230/230 [==============================] - 1s 2ms/step - loss: 0.7882 - acc: 0.7566 - val_loss: 0.8080 - val_acc: 0.7472\n",
            "Epoch 105/270\n",
            "230/230 [==============================] - 1s 2ms/step - loss: 0.7828 - acc: 0.7620 - val_loss: 0.8027 - val_acc: 0.7537\n",
            "Epoch 106/270\n",
            "230/230 [==============================] - 0s 2ms/step - loss: 0.7774 - acc: 0.7699 - val_loss: 0.7973 - val_acc: 0.7585\n",
            "Epoch 107/270\n",
            "230/230 [==============================] - 0s 2ms/step - loss: 0.7721 - acc: 0.7713 - val_loss: 0.7920 - val_acc: 0.7624\n",
            "Epoch 108/270\n",
            "230/230 [==============================] - 1s 2ms/step - loss: 0.7668 - acc: 0.7723 - val_loss: 0.7867 - val_acc: 0.7624\n",
            "Epoch 109/270\n",
            "230/230 [==============================] - 1s 2ms/step - loss: 0.7614 - acc: 0.7735 - val_loss: 0.7814 - val_acc: 0.7646\n",
            "Epoch 110/270\n",
            "230/230 [==============================] - 1s 2ms/step - loss: 0.7561 - acc: 0.7763 - val_loss: 0.7761 - val_acc: 0.7759\n",
            "Epoch 111/270\n",
            "230/230 [==============================] - 0s 2ms/step - loss: 0.7509 - acc: 0.8022 - val_loss: 0.7708 - val_acc: 0.8046\n",
            "Epoch 112/270\n",
            "230/230 [==============================] - 1s 2ms/step - loss: 0.7456 - acc: 0.8129 - val_loss: 0.7656 - val_acc: 0.8085\n",
            "Epoch 113/270\n",
            "230/230 [==============================] - 0s 2ms/step - loss: 0.7403 - acc: 0.8145 - val_loss: 0.7603 - val_acc: 0.8090\n",
            "Epoch 114/270\n",
            "230/230 [==============================] - 0s 2ms/step - loss: 0.7351 - acc: 0.8145 - val_loss: 0.7551 - val_acc: 0.8094\n",
            "Epoch 115/270\n",
            "230/230 [==============================] - 1s 2ms/step - loss: 0.7299 - acc: 0.8150 - val_loss: 0.7498 - val_acc: 0.8103\n",
            "Epoch 116/270\n",
            "230/230 [==============================] - 0s 2ms/step - loss: 0.7246 - acc: 0.8164 - val_loss: 0.7447 - val_acc: 0.8103\n",
            "Epoch 117/270\n",
            "230/230 [==============================] - 0s 2ms/step - loss: 0.7194 - acc: 0.8166 - val_loss: 0.7395 - val_acc: 0.8103\n",
            "Epoch 118/270\n",
            "230/230 [==============================] - 1s 2ms/step - loss: 0.7142 - acc: 0.8179 - val_loss: 0.7343 - val_acc: 0.8138\n",
            "Epoch 119/270\n",
            "230/230 [==============================] - 1s 2ms/step - loss: 0.7090 - acc: 0.8193 - val_loss: 0.7291 - val_acc: 0.8168\n",
            "Epoch 120/270\n",
            "230/230 [==============================] - 1s 2ms/step - loss: 0.7038 - acc: 0.8217 - val_loss: 0.7240 - val_acc: 0.8194\n",
            "Epoch 121/270\n",
            "230/230 [==============================] - 1s 2ms/step - loss: 0.6986 - acc: 0.8276 - val_loss: 0.7188 - val_acc: 0.8220\n",
            "Epoch 122/270\n",
            "230/230 [==============================] - 0s 2ms/step - loss: 0.6935 - acc: 0.8295 - val_loss: 0.7137 - val_acc: 0.8220\n",
            "Epoch 123/270\n",
            "230/230 [==============================] - 0s 2ms/step - loss: 0.6883 - acc: 0.8295 - val_loss: 0.7085 - val_acc: 0.8220\n",
            "Epoch 124/270\n",
            "230/230 [==============================] - 0s 2ms/step - loss: 0.6831 - acc: 0.8314 - val_loss: 0.7034 - val_acc: 0.8242\n",
            "Epoch 125/270\n",
            "230/230 [==============================] - 0s 2ms/step - loss: 0.6780 - acc: 0.8365 - val_loss: 0.6982 - val_acc: 0.8255\n",
            "Epoch 126/270\n",
            "230/230 [==============================] - 0s 2ms/step - loss: 0.6728 - acc: 0.8413 - val_loss: 0.6931 - val_acc: 0.8342\n",
            "Epoch 127/270\n",
            "230/230 [==============================] - 0s 2ms/step - loss: 0.6677 - acc: 0.8471 - val_loss: 0.6880 - val_acc: 0.8368\n",
            "Epoch 128/270\n",
            "230/230 [==============================] - 1s 2ms/step - loss: 0.6625 - acc: 0.8493 - val_loss: 0.6829 - val_acc: 0.8368\n",
            "Epoch 129/270\n",
            "230/230 [==============================] - 0s 2ms/step - loss: 0.6574 - acc: 0.8494 - val_loss: 0.6778 - val_acc: 0.8368\n",
            "Epoch 130/270\n",
            "230/230 [==============================] - 0s 2ms/step - loss: 0.6522 - acc: 0.8495 - val_loss: 0.6727 - val_acc: 0.8372\n",
            "Epoch 131/270\n",
            "230/230 [==============================] - 1s 2ms/step - loss: 0.6471 - acc: 0.8502 - val_loss: 0.6676 - val_acc: 0.8399\n",
            "Epoch 132/270\n",
            "230/230 [==============================] - 0s 2ms/step - loss: 0.6420 - acc: 0.8524 - val_loss: 0.6625 - val_acc: 0.8399\n",
            "Epoch 133/270\n",
            "230/230 [==============================] - 1s 2ms/step - loss: 0.6368 - acc: 0.8542 - val_loss: 0.6574 - val_acc: 0.8407\n",
            "Epoch 134/270\n",
            "230/230 [==============================] - 0s 2ms/step - loss: 0.6317 - acc: 0.8559 - val_loss: 0.6524 - val_acc: 0.8412\n",
            "Epoch 135/270\n",
            "230/230 [==============================] - 1s 2ms/step - loss: 0.6266 - acc: 0.8559 - val_loss: 0.6473 - val_acc: 0.8412\n",
            "Epoch 136/270\n",
            "230/230 [==============================] - 1s 2ms/step - loss: 0.6215 - acc: 0.8559 - val_loss: 0.6423 - val_acc: 0.8412\n",
            "Epoch 137/270\n",
            "230/230 [==============================] - 0s 2ms/step - loss: 0.6165 - acc: 0.8562 - val_loss: 0.6373 - val_acc: 0.8416\n",
            "Epoch 138/270\n",
            "230/230 [==============================] - 0s 2ms/step - loss: 0.6115 - acc: 0.8566 - val_loss: 0.6323 - val_acc: 0.8420\n",
            "Epoch 139/270\n",
            "230/230 [==============================] - 1s 2ms/step - loss: 0.6065 - acc: 0.8581 - val_loss: 0.6274 - val_acc: 0.8451\n",
            "Epoch 140/270\n",
            "230/230 [==============================] - 0s 2ms/step - loss: 0.6015 - acc: 0.8604 - val_loss: 0.6225 - val_acc: 0.8451\n",
            "Epoch 141/270\n",
            "230/230 [==============================] - 1s 2ms/step - loss: 0.5965 - acc: 0.8621 - val_loss: 0.6176 - val_acc: 0.8464\n",
            "Epoch 142/270\n",
            "230/230 [==============================] - 0s 2ms/step - loss: 0.5916 - acc: 0.8634 - val_loss: 0.6127 - val_acc: 0.8481\n",
            "Epoch 143/270\n",
            "230/230 [==============================] - 0s 2ms/step - loss: 0.5867 - acc: 0.8647 - val_loss: 0.6079 - val_acc: 0.8520\n",
            "Epoch 144/270\n",
            "230/230 [==============================] - 0s 2ms/step - loss: 0.5818 - acc: 0.8667 - val_loss: 0.6031 - val_acc: 0.8512\n",
            "Epoch 145/270\n",
            "230/230 [==============================] - 0s 2ms/step - loss: 0.5770 - acc: 0.8672 - val_loss: 0.5983 - val_acc: 0.8529\n",
            "Epoch 146/270\n",
            "230/230 [==============================] - 0s 2ms/step - loss: 0.5722 - acc: 0.8681 - val_loss: 0.5935 - val_acc: 0.8538\n",
            "Epoch 147/270\n",
            "230/230 [==============================] - 1s 2ms/step - loss: 0.5674 - acc: 0.8693 - val_loss: 0.5888 - val_acc: 0.8573\n",
            "Epoch 148/270\n",
            "230/230 [==============================] - 0s 2ms/step - loss: 0.5626 - acc: 0.8698 - val_loss: 0.5841 - val_acc: 0.8577\n",
            "Epoch 149/270\n",
            "230/230 [==============================] - 1s 2ms/step - loss: 0.5579 - acc: 0.8704 - val_loss: 0.5794 - val_acc: 0.8577\n",
            "Epoch 150/270\n",
            "230/230 [==============================] - 1s 2ms/step - loss: 0.5532 - acc: 0.8704 - val_loss: 0.5748 - val_acc: 0.8581\n",
            "Epoch 151/270\n",
            "230/230 [==============================] - 1s 2ms/step - loss: 0.5486 - acc: 0.8707 - val_loss: 0.5702 - val_acc: 0.8581\n",
            "Epoch 152/270\n",
            "230/230 [==============================] - 0s 2ms/step - loss: 0.5439 - acc: 0.8723 - val_loss: 0.5656 - val_acc: 0.8638\n",
            "Epoch 153/270\n",
            "230/230 [==============================] - 0s 2ms/step - loss: 0.5394 - acc: 0.8740 - val_loss: 0.5611 - val_acc: 0.8638\n",
            "Epoch 154/270\n",
            "230/230 [==============================] - 0s 2ms/step - loss: 0.5348 - acc: 0.8744 - val_loss: 0.5566 - val_acc: 0.8638\n",
            "Epoch 155/270\n",
            "230/230 [==============================] - 0s 2ms/step - loss: 0.5303 - acc: 0.8745 - val_loss: 0.5521 - val_acc: 0.8638\n",
            "Epoch 156/270\n",
            "230/230 [==============================] - 0s 2ms/step - loss: 0.5258 - acc: 0.8753 - val_loss: 0.5477 - val_acc: 0.8642\n",
            "Epoch 157/270\n",
            "230/230 [==============================] - 1s 2ms/step - loss: 0.5214 - acc: 0.8810 - val_loss: 0.5433 - val_acc: 0.8716\n",
            "Epoch 158/270\n",
            "230/230 [==============================] - 0s 2ms/step - loss: 0.5170 - acc: 0.8822 - val_loss: 0.5390 - val_acc: 0.8725\n",
            "Epoch 159/270\n",
            "230/230 [==============================] - 1s 2ms/step - loss: 0.5126 - acc: 0.8822 - val_loss: 0.5346 - val_acc: 0.8725\n",
            "Epoch 160/270\n",
            "230/230 [==============================] - 0s 2ms/step - loss: 0.5083 - acc: 0.8830 - val_loss: 0.5303 - val_acc: 0.8755\n",
            "Epoch 161/270\n",
            "230/230 [==============================] - 0s 2ms/step - loss: 0.5040 - acc: 0.8839 - val_loss: 0.5261 - val_acc: 0.8764\n",
            "Epoch 162/270\n",
            "230/230 [==============================] - 0s 2ms/step - loss: 0.4997 - acc: 0.8842 - val_loss: 0.5219 - val_acc: 0.8764\n",
            "Epoch 163/270\n",
            "230/230 [==============================] - 0s 2ms/step - loss: 0.4955 - acc: 0.8847 - val_loss: 0.5177 - val_acc: 0.8764\n",
            "Epoch 164/270\n",
            "230/230 [==============================] - 0s 2ms/step - loss: 0.4914 - acc: 0.8852 - val_loss: 0.5136 - val_acc: 0.8764\n",
            "Epoch 165/270\n",
            "230/230 [==============================] - 1s 2ms/step - loss: 0.4873 - acc: 0.8853 - val_loss: 0.5095 - val_acc: 0.8764\n",
            "Epoch 166/270\n",
            "230/230 [==============================] - 0s 2ms/step - loss: 0.4832 - acc: 0.8854 - val_loss: 0.5055 - val_acc: 0.8764\n",
            "Epoch 167/270\n",
            "230/230 [==============================] - 0s 2ms/step - loss: 0.4791 - acc: 0.8888 - val_loss: 0.5015 - val_acc: 0.8838\n",
            "Epoch 168/270\n",
            "230/230 [==============================] - 0s 2ms/step - loss: 0.4752 - acc: 0.8912 - val_loss: 0.4975 - val_acc: 0.8838\n",
            "Epoch 169/270\n",
            "230/230 [==============================] - 0s 2ms/step - loss: 0.4712 - acc: 0.8918 - val_loss: 0.4936 - val_acc: 0.8842\n",
            "Epoch 170/270\n",
            "230/230 [==============================] - 0s 2ms/step - loss: 0.4673 - acc: 0.8921 - val_loss: 0.4897 - val_acc: 0.8842\n",
            "Epoch 171/270\n",
            "230/230 [==============================] - 0s 2ms/step - loss: 0.4635 - acc: 0.8921 - val_loss: 0.4859 - val_acc: 0.8842\n",
            "Epoch 172/270\n",
            "230/230 [==============================] - 1s 2ms/step - loss: 0.4596 - acc: 0.8921 - val_loss: 0.4821 - val_acc: 0.8842\n",
            "Epoch 173/270\n",
            "230/230 [==============================] - 0s 2ms/step - loss: 0.4559 - acc: 0.8921 - val_loss: 0.4784 - val_acc: 0.8842\n",
            "Epoch 174/270\n",
            "230/230 [==============================] - 0s 2ms/step - loss: 0.4521 - acc: 0.8925 - val_loss: 0.4747 - val_acc: 0.8860\n",
            "Epoch 175/270\n",
            "230/230 [==============================] - 0s 2ms/step - loss: 0.4485 - acc: 0.8931 - val_loss: 0.4710 - val_acc: 0.8864\n",
            "Epoch 176/270\n",
            "230/230 [==============================] - 0s 2ms/step - loss: 0.4448 - acc: 0.8945 - val_loss: 0.4674 - val_acc: 0.8864\n",
            "Epoch 177/270\n",
            "230/230 [==============================] - 1s 2ms/step - loss: 0.4412 - acc: 0.8981 - val_loss: 0.4638 - val_acc: 0.8899\n",
            "Epoch 178/270\n",
            "230/230 [==============================] - 0s 2ms/step - loss: 0.4377 - acc: 0.8989 - val_loss: 0.4603 - val_acc: 0.8903\n",
            "Epoch 179/270\n",
            "230/230 [==============================] - 1s 2ms/step - loss: 0.4342 - acc: 0.8991 - val_loss: 0.4568 - val_acc: 0.8912\n",
            "Epoch 180/270\n",
            "230/230 [==============================] - 0s 2ms/step - loss: 0.4307 - acc: 0.9006 - val_loss: 0.4533 - val_acc: 0.8921\n",
            "Epoch 181/270\n",
            "230/230 [==============================] - 1s 2ms/step - loss: 0.4273 - acc: 0.9023 - val_loss: 0.4499 - val_acc: 0.8921\n",
            "Epoch 182/270\n",
            "230/230 [==============================] - 0s 2ms/step - loss: 0.4239 - acc: 0.9086 - val_loss: 0.4465 - val_acc: 0.8977\n",
            "Epoch 183/270\n",
            "230/230 [==============================] - 0s 2ms/step - loss: 0.4205 - acc: 0.9099 - val_loss: 0.4432 - val_acc: 0.8986\n",
            "Epoch 184/270\n",
            "230/230 [==============================] - 0s 2ms/step - loss: 0.4172 - acc: 0.9100 - val_loss: 0.4399 - val_acc: 0.8986\n",
            "Epoch 185/270\n",
            "230/230 [==============================] - 0s 2ms/step - loss: 0.4140 - acc: 0.9103 - val_loss: 0.4366 - val_acc: 0.8999\n",
            "Epoch 186/270\n",
            "230/230 [==============================] - 1s 2ms/step - loss: 0.4108 - acc: 0.9109 - val_loss: 0.4334 - val_acc: 0.8999\n",
            "Epoch 187/270\n",
            "230/230 [==============================] - 1s 2ms/step - loss: 0.4076 - acc: 0.9118 - val_loss: 0.4302 - val_acc: 0.9025\n",
            "Epoch 188/270\n",
            "230/230 [==============================] - 0s 2ms/step - loss: 0.4045 - acc: 0.9129 - val_loss: 0.4271 - val_acc: 0.9056\n",
            "Epoch 189/270\n",
            "230/230 [==============================] - 1s 2ms/step - loss: 0.4014 - acc: 0.9136 - val_loss: 0.4240 - val_acc: 0.9064\n",
            "Epoch 190/270\n",
            "230/230 [==============================] - 0s 2ms/step - loss: 0.3983 - acc: 0.9137 - val_loss: 0.4209 - val_acc: 0.9064\n",
            "Epoch 191/270\n",
            "230/230 [==============================] - 0s 2ms/step - loss: 0.3953 - acc: 0.9138 - val_loss: 0.4179 - val_acc: 0.9064\n",
            "Epoch 192/270\n",
            "230/230 [==============================] - 1s 2ms/step - loss: 0.3923 - acc: 0.9142 - val_loss: 0.4150 - val_acc: 0.9077\n",
            "Epoch 193/270\n",
            "230/230 [==============================] - 0s 2ms/step - loss: 0.3894 - acc: 0.9149 - val_loss: 0.4120 - val_acc: 0.9077\n",
            "Epoch 194/270\n",
            "230/230 [==============================] - 0s 2ms/step - loss: 0.3865 - acc: 0.9149 - val_loss: 0.4091 - val_acc: 0.9077\n",
            "Epoch 195/270\n",
            "230/230 [==============================] - 1s 2ms/step - loss: 0.3836 - acc: 0.9151 - val_loss: 0.4063 - val_acc: 0.9077\n",
            "Epoch 196/270\n",
            "230/230 [==============================] - 1s 2ms/step - loss: 0.3808 - acc: 0.9183 - val_loss: 0.4034 - val_acc: 0.9125\n",
            "Epoch 197/270\n",
            "230/230 [==============================] - 0s 2ms/step - loss: 0.3781 - acc: 0.9187 - val_loss: 0.4007 - val_acc: 0.9125\n",
            "Epoch 198/270\n",
            "230/230 [==============================] - 1s 2ms/step - loss: 0.3753 - acc: 0.9188 - val_loss: 0.3979 - val_acc: 0.9130\n",
            "Epoch 199/270\n",
            "230/230 [==============================] - 1s 2ms/step - loss: 0.3726 - acc: 0.9197 - val_loss: 0.3952 - val_acc: 0.9147\n",
            "Epoch 200/270\n",
            "230/230 [==============================] - 0s 2ms/step - loss: 0.3699 - acc: 0.9207 - val_loss: 0.3925 - val_acc: 0.9151\n",
            "Epoch 201/270\n",
            "230/230 [==============================] - 1s 2ms/step - loss: 0.3673 - acc: 0.9209 - val_loss: 0.3899 - val_acc: 0.9151\n",
            "Epoch 202/270\n",
            "230/230 [==============================] - 1s 2ms/step - loss: 0.3647 - acc: 0.9209 - val_loss: 0.3873 - val_acc: 0.9151\n",
            "Epoch 203/270\n",
            "230/230 [==============================] - 0s 2ms/step - loss: 0.3621 - acc: 0.9212 - val_loss: 0.3847 - val_acc: 0.9151\n",
            "Epoch 204/270\n",
            "230/230 [==============================] - 0s 2ms/step - loss: 0.3596 - acc: 0.9219 - val_loss: 0.3821 - val_acc: 0.9151\n",
            "Epoch 205/270\n",
            "230/230 [==============================] - 1s 2ms/step - loss: 0.3571 - acc: 0.9216 - val_loss: 0.3796 - val_acc: 0.9169\n",
            "Epoch 206/270\n",
            "230/230 [==============================] - 1s 2ms/step - loss: 0.3547 - acc: 0.9227 - val_loss: 0.3771 - val_acc: 0.9169\n",
            "Epoch 207/270\n",
            "230/230 [==============================] - 0s 2ms/step - loss: 0.3522 - acc: 0.9231 - val_loss: 0.3747 - val_acc: 0.9169\n",
            "Epoch 208/270\n",
            "230/230 [==============================] - 0s 2ms/step - loss: 0.3498 - acc: 0.9231 - val_loss: 0.3723 - val_acc: 0.9169\n",
            "Epoch 209/270\n",
            "230/230 [==============================] - 0s 2ms/step - loss: 0.3475 - acc: 0.9232 - val_loss: 0.3699 - val_acc: 0.9169\n",
            "Epoch 210/270\n",
            "230/230 [==============================] - 0s 2ms/step - loss: 0.3451 - acc: 0.9232 - val_loss: 0.3676 - val_acc: 0.9169\n",
            "Epoch 211/270\n",
            "230/230 [==============================] - 1s 2ms/step - loss: 0.3428 - acc: 0.9240 - val_loss: 0.3652 - val_acc: 0.9182\n",
            "Epoch 212/270\n",
            "230/230 [==============================] - 1s 2ms/step - loss: 0.3406 - acc: 0.9245 - val_loss: 0.3629 - val_acc: 0.9182\n",
            "Epoch 213/270\n",
            "230/230 [==============================] - 1s 2ms/step - loss: 0.3383 - acc: 0.9247 - val_loss: 0.3607 - val_acc: 0.9182\n",
            "Epoch 214/270\n",
            "230/230 [==============================] - 1s 2ms/step - loss: 0.3361 - acc: 0.9251 - val_loss: 0.3585 - val_acc: 0.9182\n",
            "Epoch 215/270\n",
            "230/230 [==============================] - 1s 2ms/step - loss: 0.3339 - acc: 0.9259 - val_loss: 0.3563 - val_acc: 0.9182\n",
            "Epoch 216/270\n",
            "230/230 [==============================] - 1s 2ms/step - loss: 0.3318 - acc: 0.9277 - val_loss: 0.3541 - val_acc: 0.9221\n",
            "Epoch 217/270\n",
            "230/230 [==============================] - 0s 2ms/step - loss: 0.3297 - acc: 0.9293 - val_loss: 0.3519 - val_acc: 0.9221\n",
            "Epoch 218/270\n",
            "230/230 [==============================] - 1s 2ms/step - loss: 0.3276 - acc: 0.9294 - val_loss: 0.3498 - val_acc: 0.9221\n",
            "Epoch 219/270\n",
            "230/230 [==============================] - 1s 2ms/step - loss: 0.3255 - acc: 0.9293 - val_loss: 0.3477 - val_acc: 0.9221\n",
            "Epoch 220/270\n",
            "230/230 [==============================] - 0s 2ms/step - loss: 0.3235 - acc: 0.9291 - val_loss: 0.3457 - val_acc: 0.9221\n",
            "Epoch 221/270\n",
            "230/230 [==============================] - 1s 2ms/step - loss: 0.3214 - acc: 0.9294 - val_loss: 0.3436 - val_acc: 0.9221\n",
            "Epoch 222/270\n",
            "230/230 [==============================] - 1s 2ms/step - loss: 0.3195 - acc: 0.9294 - val_loss: 0.3416 - val_acc: 0.9221\n",
            "Epoch 223/270\n",
            "230/230 [==============================] - 1s 2ms/step - loss: 0.3175 - acc: 0.9294 - val_loss: 0.3397 - val_acc: 0.9221\n",
            "Epoch 224/270\n",
            "230/230 [==============================] - 1s 2ms/step - loss: 0.3156 - acc: 0.9294 - val_loss: 0.3377 - val_acc: 0.9221\n",
            "Epoch 225/270\n",
            "230/230 [==============================] - 1s 2ms/step - loss: 0.3137 - acc: 0.9297 - val_loss: 0.3358 - val_acc: 0.9230\n",
            "Epoch 226/270\n",
            "230/230 [==============================] - 1s 2ms/step - loss: 0.3118 - acc: 0.9298 - val_loss: 0.3339 - val_acc: 0.9234\n",
            "Epoch 227/270\n",
            "230/230 [==============================] - 1s 2ms/step - loss: 0.3099 - acc: 0.9313 - val_loss: 0.3320 - val_acc: 0.9234\n",
            "Epoch 228/270\n",
            "230/230 [==============================] - 1s 2ms/step - loss: 0.3081 - acc: 0.9318 - val_loss: 0.3301 - val_acc: 0.9234\n",
            "Epoch 229/270\n",
            "230/230 [==============================] - 1s 2ms/step - loss: 0.3062 - acc: 0.9318 - val_loss: 0.3283 - val_acc: 0.9234\n",
            "Epoch 230/270\n",
            "230/230 [==============================] - 1s 2ms/step - loss: 0.3044 - acc: 0.9318 - val_loss: 0.3265 - val_acc: 0.9273\n",
            "Epoch 231/270\n",
            "230/230 [==============================] - 1s 2ms/step - loss: 0.3027 - acc: 0.9337 - val_loss: 0.3247 - val_acc: 0.9286\n",
            "Epoch 232/270\n",
            "230/230 [==============================] - 1s 2ms/step - loss: 0.3009 - acc: 0.9358 - val_loss: 0.3229 - val_acc: 0.9291\n",
            "Epoch 233/270\n",
            "230/230 [==============================] - 0s 2ms/step - loss: 0.2992 - acc: 0.9356 - val_loss: 0.3212 - val_acc: 0.9291\n",
            "Epoch 234/270\n",
            "230/230 [==============================] - 0s 2ms/step - loss: 0.2975 - acc: 0.9360 - val_loss: 0.3195 - val_acc: 0.9291\n",
            "Epoch 235/270\n",
            "230/230 [==============================] - 1s 2ms/step - loss: 0.2958 - acc: 0.9364 - val_loss: 0.3178 - val_acc: 0.9295\n",
            "Epoch 236/270\n",
            "230/230 [==============================] - 1s 2ms/step - loss: 0.2941 - acc: 0.9370 - val_loss: 0.3161 - val_acc: 0.9304\n",
            "Epoch 237/270\n",
            "230/230 [==============================] - 1s 2ms/step - loss: 0.2925 - acc: 0.9374 - val_loss: 0.3144 - val_acc: 0.9308\n",
            "Epoch 238/270\n",
            "230/230 [==============================] - 0s 2ms/step - loss: 0.2909 - acc: 0.9384 - val_loss: 0.3128 - val_acc: 0.9304\n",
            "Epoch 239/270\n",
            "230/230 [==============================] - 0s 2ms/step - loss: 0.2893 - acc: 0.9388 - val_loss: 0.3112 - val_acc: 0.9321\n",
            "Epoch 240/270\n",
            "230/230 [==============================] - 0s 2ms/step - loss: 0.2877 - acc: 0.9399 - val_loss: 0.3095 - val_acc: 0.9321\n",
            "Epoch 241/270\n",
            "230/230 [==============================] - 0s 2ms/step - loss: 0.2861 - acc: 0.9400 - val_loss: 0.3080 - val_acc: 0.9326\n",
            "Epoch 242/270\n",
            "230/230 [==============================] - 1s 2ms/step - loss: 0.2845 - acc: 0.9407 - val_loss: 0.3064 - val_acc: 0.9326\n",
            "Epoch 243/270\n",
            "230/230 [==============================] - 0s 2ms/step - loss: 0.2830 - acc: 0.9404 - val_loss: 0.3048 - val_acc: 0.9343\n",
            "Epoch 244/270\n",
            "230/230 [==============================] - 0s 2ms/step - loss: 0.2815 - acc: 0.9417 - val_loss: 0.3033 - val_acc: 0.9347\n",
            "Epoch 245/270\n",
            "230/230 [==============================] - 0s 2ms/step - loss: 0.2800 - acc: 0.9417 - val_loss: 0.3018 - val_acc: 0.9347\n",
            "Epoch 246/270\n",
            "230/230 [==============================] - 1s 2ms/step - loss: 0.2785 - acc: 0.9417 - val_loss: 0.3003 - val_acc: 0.9352\n",
            "Epoch 247/270\n",
            "230/230 [==============================] - 0s 2ms/step - loss: 0.2771 - acc: 0.9417 - val_loss: 0.2988 - val_acc: 0.9352\n",
            "Epoch 248/270\n",
            "230/230 [==============================] - 0s 2ms/step - loss: 0.2756 - acc: 0.9417 - val_loss: 0.2973 - val_acc: 0.9352\n",
            "Epoch 249/270\n",
            "230/230 [==============================] - 0s 2ms/step - loss: 0.2742 - acc: 0.9419 - val_loss: 0.2959 - val_acc: 0.9352\n",
            "Epoch 250/270\n",
            "230/230 [==============================] - 1s 2ms/step - loss: 0.2728 - acc: 0.9423 - val_loss: 0.2945 - val_acc: 0.9352\n",
            "Epoch 251/270\n",
            "230/230 [==============================] - 1s 2ms/step - loss: 0.2714 - acc: 0.9422 - val_loss: 0.2931 - val_acc: 0.9352\n",
            "Epoch 252/270\n",
            "230/230 [==============================] - 1s 2ms/step - loss: 0.2700 - acc: 0.9424 - val_loss: 0.2917 - val_acc: 0.9352\n",
            "Epoch 253/270\n",
            "230/230 [==============================] - 1s 2ms/step - loss: 0.2686 - acc: 0.9426 - val_loss: 0.2903 - val_acc: 0.9352\n",
            "Epoch 254/270\n",
            "230/230 [==============================] - 0s 2ms/step - loss: 0.2673 - acc: 0.9424 - val_loss: 0.2889 - val_acc: 0.9343\n",
            "Epoch 255/270\n",
            "230/230 [==============================] - 0s 2ms/step - loss: 0.2659 - acc: 0.9424 - val_loss: 0.2876 - val_acc: 0.9343\n",
            "Epoch 256/270\n",
            "230/230 [==============================] - 1s 2ms/step - loss: 0.2646 - acc: 0.9425 - val_loss: 0.2862 - val_acc: 0.9347\n",
            "Epoch 257/270\n",
            "230/230 [==============================] - 0s 2ms/step - loss: 0.2633 - acc: 0.9426 - val_loss: 0.2849 - val_acc: 0.9347\n",
            "Epoch 258/270\n",
            "230/230 [==============================] - 0s 2ms/step - loss: 0.2620 - acc: 0.9429 - val_loss: 0.2836 - val_acc: 0.9347\n",
            "Epoch 259/270\n",
            "230/230 [==============================] - 1s 2ms/step - loss: 0.2607 - acc: 0.9431 - val_loss: 0.2823 - val_acc: 0.9347\n",
            "Epoch 260/270\n",
            "230/230 [==============================] - 1s 2ms/step - loss: 0.2595 - acc: 0.9437 - val_loss: 0.2810 - val_acc: 0.9356\n",
            "Epoch 261/270\n",
            "230/230 [==============================] - 0s 2ms/step - loss: 0.2582 - acc: 0.9442 - val_loss: 0.2798 - val_acc: 0.9356\n",
            "Epoch 262/270\n",
            "230/230 [==============================] - 1s 2ms/step - loss: 0.2570 - acc: 0.9442 - val_loss: 0.2785 - val_acc: 0.9356\n",
            "Epoch 263/270\n",
            "230/230 [==============================] - 0s 2ms/step - loss: 0.2557 - acc: 0.9443 - val_loss: 0.2772 - val_acc: 0.9356\n",
            "Epoch 264/270\n",
            "230/230 [==============================] - 1s 2ms/step - loss: 0.2545 - acc: 0.9447 - val_loss: 0.2760 - val_acc: 0.9356\n",
            "Epoch 265/270\n",
            "230/230 [==============================] - 0s 2ms/step - loss: 0.2533 - acc: 0.9456 - val_loss: 0.2748 - val_acc: 0.9356\n",
            "Epoch 266/270\n",
            "230/230 [==============================] - 0s 2ms/step - loss: 0.2521 - acc: 0.9453 - val_loss: 0.2736 - val_acc: 0.9360\n",
            "Epoch 267/270\n",
            "230/230 [==============================] - 0s 2ms/step - loss: 0.2509 - acc: 0.9468 - val_loss: 0.2724 - val_acc: 0.9360\n",
            "Epoch 268/270\n",
            "230/230 [==============================] - 0s 2ms/step - loss: 0.2498 - acc: 0.9463 - val_loss: 0.2712 - val_acc: 0.9360\n",
            "Epoch 269/270\n",
            "230/230 [==============================] - 0s 2ms/step - loss: 0.2486 - acc: 0.9467 - val_loss: 0.2700 - val_acc: 0.9365\n",
            "Epoch 270/270\n",
            "230/230 [==============================] - 0s 2ms/step - loss: 0.2475 - acc: 0.9467 - val_loss: 0.2689 - val_acc: 0.9365\n",
            "<bound method Model.summary of <tensorflow.python.keras.engine.sequential.Sequential object at 0x7fd8437b5e48>>\n",
            "Test Accuracy:\n",
            "2585/2585 [==============================] - 4s 2ms/step - loss: 0.2527 - acc: 0.9437\n",
            "Random Sample Accuracy:\n",
            "38/38 [==============================] - 0s 2ms/step - loss: 0.8560 - acc: 0.7193\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 1080x432 with 2 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "HN_AYIMG7RXp"
      },
      "source": [
        "model6_params = {'activation 1': 'relu', \n",
        "                 'activation 2': 'relu', \n",
        "                 'activation 3': 'softmax', \n",
        "                 'opt': tf.keras.optimizers.SGD(learning_rate=0.001),\n",
        "                 'loss': 'categorical_crossentropy',\n",
        "                 'epochs': 170, \n",
        "                 'batch_size': 30, \n",
        "                 }"
      ],
      "execution_count": 50,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 1000
        },
        "id": "brxftL147ua1",
        "outputId": "96b849d3-1f79-49a0-df22-a8acef09b764"
      },
      "source": [
        "model6_history, model6 = create_model(train, label_train, test, label_test, model6_params)"
      ],
      "execution_count": 51,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Epoch 1/170\n",
            "307/307 [==============================] - 1s 2ms/step - loss: 2.5912 - acc: 0.3847 - val_loss: 2.5368 - val_acc: 0.5117\n",
            "Epoch 2/170\n",
            "307/307 [==============================] - 1s 2ms/step - loss: 2.4790 - acc: 0.5215 - val_loss: 2.4241 - val_acc: 0.5109\n",
            "Epoch 3/170\n",
            "307/307 [==============================] - 1s 2ms/step - loss: 2.3600 - acc: 0.5215 - val_loss: 2.3012 - val_acc: 0.5109\n",
            "Epoch 4/170\n",
            "307/307 [==============================] - 1s 2ms/step - loss: 2.2257 - acc: 0.5215 - val_loss: 2.1648 - val_acc: 0.5109\n",
            "Epoch 5/170\n",
            "307/307 [==============================] - 1s 2ms/step - loss: 2.0840 - acc: 0.5215 - val_loss: 2.0258 - val_acc: 0.5109\n",
            "Epoch 6/170\n",
            "307/307 [==============================] - 1s 2ms/step - loss: 1.9439 - acc: 0.5215 - val_loss: 1.8952 - val_acc: 0.5109\n",
            "Epoch 7/170\n",
            "307/307 [==============================] - 1s 2ms/step - loss: 1.8204 - acc: 0.5215 - val_loss: 1.7892 - val_acc: 0.5109\n",
            "Epoch 8/170\n",
            "307/307 [==============================] - 1s 2ms/step - loss: 1.7269 - acc: 0.5215 - val_loss: 1.7154 - val_acc: 0.5109\n",
            "Epoch 9/170\n",
            "307/307 [==============================] - 1s 2ms/step - loss: 1.6639 - acc: 0.5215 - val_loss: 1.6656 - val_acc: 0.5109\n",
            "Epoch 10/170\n",
            "307/307 [==============================] - 1s 2ms/step - loss: 1.6193 - acc: 0.5215 - val_loss: 1.6278 - val_acc: 0.5109\n",
            "Epoch 11/170\n",
            "307/307 [==============================] - 1s 2ms/step - loss: 1.5842 - acc: 0.5215 - val_loss: 1.5965 - val_acc: 0.5109\n",
            "Epoch 12/170\n",
            "307/307 [==============================] - 1s 2ms/step - loss: 1.5544 - acc: 0.5215 - val_loss: 1.5693 - val_acc: 0.5109\n",
            "Epoch 13/170\n",
            "307/307 [==============================] - 1s 2ms/step - loss: 1.5281 - acc: 0.5215 - val_loss: 1.5444 - val_acc: 0.5109\n",
            "Epoch 14/170\n",
            "307/307 [==============================] - 1s 2ms/step - loss: 1.5038 - acc: 0.5215 - val_loss: 1.5210 - val_acc: 0.5109\n",
            "Epoch 15/170\n",
            "307/307 [==============================] - 1s 2ms/step - loss: 1.4807 - acc: 0.5215 - val_loss: 1.4985 - val_acc: 0.5109\n",
            "Epoch 16/170\n",
            "307/307 [==============================] - 1s 2ms/step - loss: 1.4586 - acc: 0.5215 - val_loss: 1.4768 - val_acc: 0.5109\n",
            "Epoch 17/170\n",
            "307/307 [==============================] - 1s 2ms/step - loss: 1.4373 - acc: 0.5215 - val_loss: 1.4561 - val_acc: 0.5109\n",
            "Epoch 18/170\n",
            "307/307 [==============================] - 1s 2ms/step - loss: 1.4169 - acc: 0.5215 - val_loss: 1.4360 - val_acc: 0.5109\n",
            "Epoch 19/170\n",
            "307/307 [==============================] - 1s 2ms/step - loss: 1.3970 - acc: 0.5215 - val_loss: 1.4161 - val_acc: 0.5109\n",
            "Epoch 20/170\n",
            "307/307 [==============================] - 1s 2ms/step - loss: 1.3776 - acc: 0.5215 - val_loss: 1.3967 - val_acc: 0.5109\n",
            "Epoch 21/170\n",
            "307/307 [==============================] - 1s 2ms/step - loss: 1.3585 - acc: 0.5215 - val_loss: 1.3777 - val_acc: 0.5109\n",
            "Epoch 22/170\n",
            "307/307 [==============================] - 1s 2ms/step - loss: 1.3402 - acc: 0.5215 - val_loss: 1.3597 - val_acc: 0.5109\n",
            "Epoch 23/170\n",
            "307/307 [==============================] - 1s 2ms/step - loss: 1.3226 - acc: 0.5215 - val_loss: 1.3421 - val_acc: 0.5109\n",
            "Epoch 24/170\n",
            "307/307 [==============================] - 1s 2ms/step - loss: 1.3055 - acc: 0.5215 - val_loss: 1.3250 - val_acc: 0.5109\n",
            "Epoch 25/170\n",
            "307/307 [==============================] - 1s 2ms/step - loss: 1.2888 - acc: 0.5215 - val_loss: 1.3084 - val_acc: 0.5109\n",
            "Epoch 26/170\n",
            "307/307 [==============================] - 1s 2ms/step - loss: 1.2726 - acc: 0.5215 - val_loss: 1.2922 - val_acc: 0.5109\n",
            "Epoch 27/170\n",
            "307/307 [==============================] - 1s 2ms/step - loss: 1.2570 - acc: 0.5274 - val_loss: 1.2766 - val_acc: 0.5444\n",
            "Epoch 28/170\n",
            "307/307 [==============================] - 1s 2ms/step - loss: 1.2417 - acc: 0.5723 - val_loss: 1.2613 - val_acc: 0.5679\n",
            "Epoch 29/170\n",
            "307/307 [==============================] - 1s 2ms/step - loss: 1.2270 - acc: 0.5735 - val_loss: 1.2466 - val_acc: 0.5679\n",
            "Epoch 30/170\n",
            "307/307 [==============================] - 1s 2ms/step - loss: 1.2126 - acc: 0.5735 - val_loss: 1.2322 - val_acc: 0.5679\n",
            "Epoch 31/170\n",
            "307/307 [==============================] - 1s 2ms/step - loss: 1.1986 - acc: 0.5737 - val_loss: 1.2181 - val_acc: 0.5683\n",
            "Epoch 32/170\n",
            "307/307 [==============================] - 1s 2ms/step - loss: 1.1850 - acc: 0.5797 - val_loss: 1.2044 - val_acc: 0.5762\n",
            "Epoch 33/170\n",
            "307/307 [==============================] - 1s 2ms/step - loss: 1.1717 - acc: 0.5897 - val_loss: 1.1910 - val_acc: 0.5801\n",
            "Epoch 34/170\n",
            "307/307 [==============================] - 1s 2ms/step - loss: 1.1587 - acc: 0.5911 - val_loss: 1.1781 - val_acc: 0.5831\n",
            "Epoch 35/170\n",
            "307/307 [==============================] - 1s 2ms/step - loss: 1.1461 - acc: 0.5976 - val_loss: 1.1654 - val_acc: 0.5905\n",
            "Epoch 36/170\n",
            "307/307 [==============================] - 1s 2ms/step - loss: 1.1339 - acc: 0.6071 - val_loss: 1.1531 - val_acc: 0.6044\n",
            "Epoch 37/170\n",
            "307/307 [==============================] - 1s 2ms/step - loss: 1.1220 - acc: 0.6191 - val_loss: 1.1412 - val_acc: 0.6140\n",
            "Epoch 38/170\n",
            "307/307 [==============================] - 1s 2ms/step - loss: 1.1104 - acc: 0.6246 - val_loss: 1.1296 - val_acc: 0.6214\n",
            "Epoch 39/170\n",
            "307/307 [==============================] - 1s 2ms/step - loss: 1.0992 - acc: 0.6351 - val_loss: 1.1183 - val_acc: 0.6301\n",
            "Epoch 40/170\n",
            "307/307 [==============================] - 1s 2ms/step - loss: 1.0882 - acc: 0.6508 - val_loss: 1.1073 - val_acc: 0.6432\n",
            "Epoch 41/170\n",
            "307/307 [==============================] - 1s 2ms/step - loss: 1.0776 - acc: 0.6688 - val_loss: 1.0966 - val_acc: 0.6936\n",
            "Epoch 42/170\n",
            "307/307 [==============================] - 1s 2ms/step - loss: 1.0672 - acc: 0.6998 - val_loss: 1.0861 - val_acc: 0.6941\n",
            "Epoch 43/170\n",
            "307/307 [==============================] - 1s 2ms/step - loss: 1.0570 - acc: 0.7002 - val_loss: 1.0759 - val_acc: 0.6967\n",
            "Epoch 44/170\n",
            "307/307 [==============================] - 1s 2ms/step - loss: 1.0471 - acc: 0.7007 - val_loss: 1.0659 - val_acc: 0.6976\n",
            "Epoch 45/170\n",
            "307/307 [==============================] - 1s 2ms/step - loss: 1.0374 - acc: 0.7012 - val_loss: 1.0561 - val_acc: 0.6963\n",
            "Epoch 46/170\n",
            "307/307 [==============================] - 1s 2ms/step - loss: 1.0278 - acc: 0.7042 - val_loss: 1.0464 - val_acc: 0.7032\n",
            "Epoch 47/170\n",
            "307/307 [==============================] - 1s 2ms/step - loss: 1.0184 - acc: 0.7101 - val_loss: 1.0369 - val_acc: 0.7097\n",
            "Epoch 48/170\n",
            "307/307 [==============================] - 1s 2ms/step - loss: 1.0092 - acc: 0.7167 - val_loss: 1.0276 - val_acc: 0.7167\n",
            "Epoch 49/170\n",
            "307/307 [==============================] - 1s 2ms/step - loss: 1.0001 - acc: 0.7176 - val_loss: 1.0184 - val_acc: 0.7167\n",
            "Epoch 50/170\n",
            "307/307 [==============================] - 1s 2ms/step - loss: 0.9912 - acc: 0.7184 - val_loss: 1.0094 - val_acc: 0.7185\n",
            "Epoch 51/170\n",
            "307/307 [==============================] - 1s 2ms/step - loss: 0.9824 - acc: 0.7193 - val_loss: 1.0005 - val_acc: 0.7215\n",
            "Epoch 52/170\n",
            "307/307 [==============================] - 1s 2ms/step - loss: 0.9737 - acc: 0.7213 - val_loss: 0.9916 - val_acc: 0.7215\n",
            "Epoch 53/170\n",
            "307/307 [==============================] - 1s 2ms/step - loss: 0.9651 - acc: 0.7232 - val_loss: 0.9829 - val_acc: 0.7237\n",
            "Epoch 54/170\n",
            "307/307 [==============================] - 1s 2ms/step - loss: 0.9566 - acc: 0.7240 - val_loss: 0.9743 - val_acc: 0.7237\n",
            "Epoch 55/170\n",
            "307/307 [==============================] - 1s 2ms/step - loss: 0.9482 - acc: 0.7239 - val_loss: 0.9657 - val_acc: 0.7237\n",
            "Epoch 56/170\n",
            "307/307 [==============================] - 1s 2ms/step - loss: 0.9398 - acc: 0.7267 - val_loss: 0.9572 - val_acc: 0.7267\n",
            "Epoch 57/170\n",
            "307/307 [==============================] - 1s 2ms/step - loss: 0.9315 - acc: 0.7317 - val_loss: 0.9488 - val_acc: 0.7293\n",
            "Epoch 58/170\n",
            "307/307 [==============================] - 1s 2ms/step - loss: 0.9233 - acc: 0.7348 - val_loss: 0.9405 - val_acc: 0.7341\n",
            "Epoch 59/170\n",
            "307/307 [==============================] - 1s 2ms/step - loss: 0.9152 - acc: 0.7405 - val_loss: 0.9323 - val_acc: 0.7402\n",
            "Epoch 60/170\n",
            "307/307 [==============================] - 1s 2ms/step - loss: 0.9071 - acc: 0.7439 - val_loss: 0.9241 - val_acc: 0.7402\n",
            "Epoch 61/170\n",
            "307/307 [==============================] - 1s 2ms/step - loss: 0.8991 - acc: 0.7439 - val_loss: 0.9160 - val_acc: 0.7402\n",
            "Epoch 62/170\n",
            "307/307 [==============================] - 1s 2ms/step - loss: 0.8911 - acc: 0.7458 - val_loss: 0.9079 - val_acc: 0.7459\n",
            "Epoch 63/170\n",
            "307/307 [==============================] - 1s 2ms/step - loss: 0.8832 - acc: 0.7484 - val_loss: 0.8998 - val_acc: 0.7459\n",
            "Epoch 64/170\n",
            "307/307 [==============================] - 1s 2ms/step - loss: 0.8753 - acc: 0.7484 - val_loss: 0.8918 - val_acc: 0.7428\n",
            "Epoch 65/170\n",
            "307/307 [==============================] - 1s 2ms/step - loss: 0.8674 - acc: 0.7495 - val_loss: 0.8839 - val_acc: 0.7454\n",
            "Epoch 66/170\n",
            "307/307 [==============================] - 1s 2ms/step - loss: 0.8596 - acc: 0.7545 - val_loss: 0.8761 - val_acc: 0.7507\n",
            "Epoch 67/170\n",
            "307/307 [==============================] - 1s 2ms/step - loss: 0.8519 - acc: 0.7597 - val_loss: 0.8683 - val_acc: 0.7511\n",
            "Epoch 68/170\n",
            "307/307 [==============================] - 1s 2ms/step - loss: 0.8442 - acc: 0.7616 - val_loss: 0.8605 - val_acc: 0.7541\n",
            "Epoch 69/170\n",
            "307/307 [==============================] - 1s 2ms/step - loss: 0.8366 - acc: 0.7633 - val_loss: 0.8529 - val_acc: 0.7554\n",
            "Epoch 70/170\n",
            "307/307 [==============================] - 1s 2ms/step - loss: 0.8291 - acc: 0.7633 - val_loss: 0.8454 - val_acc: 0.7554\n",
            "Epoch 71/170\n",
            "307/307 [==============================] - 1s 2ms/step - loss: 0.8217 - acc: 0.7635 - val_loss: 0.8379 - val_acc: 0.7572\n",
            "Epoch 72/170\n",
            "307/307 [==============================] - 1s 2ms/step - loss: 0.8143 - acc: 0.7649 - val_loss: 0.8305 - val_acc: 0.7572\n",
            "Epoch 73/170\n",
            "307/307 [==============================] - 1s 2ms/step - loss: 0.8069 - acc: 0.7659 - val_loss: 0.8232 - val_acc: 0.7598\n",
            "Epoch 74/170\n",
            "307/307 [==============================] - 1s 2ms/step - loss: 0.7996 - acc: 0.7695 - val_loss: 0.8159 - val_acc: 0.7598\n",
            "Epoch 75/170\n",
            "307/307 [==============================] - 1s 2ms/step - loss: 0.7923 - acc: 0.7696 - val_loss: 0.8086 - val_acc: 0.7607\n",
            "Epoch 76/170\n",
            "307/307 [==============================] - 1s 2ms/step - loss: 0.7850 - acc: 0.7710 - val_loss: 0.8014 - val_acc: 0.7633\n",
            "Epoch 77/170\n",
            "307/307 [==============================] - 1s 2ms/step - loss: 0.7778 - acc: 0.7748 - val_loss: 0.7942 - val_acc: 0.7681\n",
            "Epoch 78/170\n",
            "307/307 [==============================] - 1s 2ms/step - loss: 0.7706 - acc: 0.7769 - val_loss: 0.7871 - val_acc: 0.7681\n",
            "Epoch 79/170\n",
            "307/307 [==============================] - 1s 2ms/step - loss: 0.7635 - acc: 0.7769 - val_loss: 0.7800 - val_acc: 0.7681\n",
            "Epoch 80/170\n",
            "307/307 [==============================] - 1s 2ms/step - loss: 0.7564 - acc: 0.7772 - val_loss: 0.7729 - val_acc: 0.7694\n",
            "Epoch 81/170\n",
            "307/307 [==============================] - 1s 2ms/step - loss: 0.7493 - acc: 0.7780 - val_loss: 0.7659 - val_acc: 0.7702\n",
            "Epoch 82/170\n",
            "307/307 [==============================] - 1s 2ms/step - loss: 0.7422 - acc: 0.7822 - val_loss: 0.7589 - val_acc: 0.7746\n",
            "Epoch 83/170\n",
            "307/307 [==============================] - 1s 2ms/step - loss: 0.7352 - acc: 0.7853 - val_loss: 0.7520 - val_acc: 0.7746\n",
            "Epoch 84/170\n",
            "307/307 [==============================] - 1s 2ms/step - loss: 0.7282 - acc: 0.7856 - val_loss: 0.7450 - val_acc: 0.7755\n",
            "Epoch 85/170\n",
            "307/307 [==============================] - 1s 2ms/step - loss: 0.7212 - acc: 0.7868 - val_loss: 0.7382 - val_acc: 0.7755\n",
            "Epoch 86/170\n",
            "307/307 [==============================] - 1s 2ms/step - loss: 0.7143 - acc: 0.7872 - val_loss: 0.7313 - val_acc: 0.7846\n",
            "Epoch 87/170\n",
            "307/307 [==============================] - 1s 2ms/step - loss: 0.7074 - acc: 0.8012 - val_loss: 0.7245 - val_acc: 0.7929\n",
            "Epoch 88/170\n",
            "307/307 [==============================] - 1s 2ms/step - loss: 0.7005 - acc: 0.8040 - val_loss: 0.7177 - val_acc: 0.7933\n",
            "Epoch 89/170\n",
            "307/307 [==============================] - 1s 2ms/step - loss: 0.6936 - acc: 0.8065 - val_loss: 0.7110 - val_acc: 0.7972\n",
            "Epoch 90/170\n",
            "307/307 [==============================] - 1s 2ms/step - loss: 0.6868 - acc: 0.8091 - val_loss: 0.7042 - val_acc: 0.7990\n",
            "Epoch 91/170\n",
            "307/307 [==============================] - 1s 2ms/step - loss: 0.6800 - acc: 0.8106 - val_loss: 0.6975 - val_acc: 0.8016\n",
            "Epoch 92/170\n",
            "307/307 [==============================] - 1s 2ms/step - loss: 0.6733 - acc: 0.8131 - val_loss: 0.6909 - val_acc: 0.8016\n",
            "Epoch 93/170\n",
            "307/307 [==============================] - 1s 2ms/step - loss: 0.6666 - acc: 0.8217 - val_loss: 0.6843 - val_acc: 0.8211\n",
            "Epoch 94/170\n",
            "307/307 [==============================] - 1s 2ms/step - loss: 0.6599 - acc: 0.8310 - val_loss: 0.6777 - val_acc: 0.8216\n",
            "Epoch 95/170\n",
            "307/307 [==============================] - 1s 2ms/step - loss: 0.6532 - acc: 0.8334 - val_loss: 0.6711 - val_acc: 0.8294\n",
            "Epoch 96/170\n",
            "307/307 [==============================] - 1s 2ms/step - loss: 0.6466 - acc: 0.8415 - val_loss: 0.6645 - val_acc: 0.8399\n",
            "Epoch 97/170\n",
            "307/307 [==============================] - 1s 2ms/step - loss: 0.6400 - acc: 0.8465 - val_loss: 0.6580 - val_acc: 0.8429\n",
            "Epoch 98/170\n",
            "307/307 [==============================] - 1s 2ms/step - loss: 0.6334 - acc: 0.8497 - val_loss: 0.6515 - val_acc: 0.8464\n",
            "Epoch 99/170\n",
            "307/307 [==============================] - 1s 2ms/step - loss: 0.6269 - acc: 0.8508 - val_loss: 0.6451 - val_acc: 0.8464\n",
            "Epoch 100/170\n",
            "307/307 [==============================] - 1s 2ms/step - loss: 0.6204 - acc: 0.8505 - val_loss: 0.6387 - val_acc: 0.8473\n",
            "Epoch 101/170\n",
            "307/307 [==============================] - 1s 2ms/step - loss: 0.6139 - acc: 0.8563 - val_loss: 0.6323 - val_acc: 0.8512\n",
            "Epoch 102/170\n",
            "307/307 [==============================] - 1s 2ms/step - loss: 0.6075 - acc: 0.8571 - val_loss: 0.6259 - val_acc: 0.8512\n",
            "Epoch 103/170\n",
            "307/307 [==============================] - 1s 2ms/step - loss: 0.6011 - acc: 0.8577 - val_loss: 0.6197 - val_acc: 0.8516\n",
            "Epoch 104/170\n",
            "307/307 [==============================] - 1s 2ms/step - loss: 0.5947 - acc: 0.8598 - val_loss: 0.6133 - val_acc: 0.8555\n",
            "Epoch 105/170\n",
            "307/307 [==============================] - 1s 2ms/step - loss: 0.5884 - acc: 0.8622 - val_loss: 0.6071 - val_acc: 0.8560\n",
            "Epoch 106/170\n",
            "307/307 [==============================] - 1s 2ms/step - loss: 0.5822 - acc: 0.8643 - val_loss: 0.6009 - val_acc: 0.8586\n",
            "Epoch 107/170\n",
            "307/307 [==============================] - 1s 2ms/step - loss: 0.5760 - acc: 0.8659 - val_loss: 0.5948 - val_acc: 0.8586\n",
            "Epoch 108/170\n",
            "307/307 [==============================] - 1s 2ms/step - loss: 0.5698 - acc: 0.8659 - val_loss: 0.5887 - val_acc: 0.8586\n",
            "Epoch 109/170\n",
            "307/307 [==============================] - 1s 2ms/step - loss: 0.5637 - acc: 0.8659 - val_loss: 0.5826 - val_acc: 0.8586\n",
            "Epoch 110/170\n",
            "307/307 [==============================] - 1s 2ms/step - loss: 0.5576 - acc: 0.8659 - val_loss: 0.5765 - val_acc: 0.8607\n",
            "Epoch 111/170\n",
            "307/307 [==============================] - 1s 2ms/step - loss: 0.5516 - acc: 0.8677 - val_loss: 0.5705 - val_acc: 0.8607\n",
            "Epoch 112/170\n",
            "307/307 [==============================] - 1s 2ms/step - loss: 0.5456 - acc: 0.8677 - val_loss: 0.5647 - val_acc: 0.8607\n",
            "Epoch 113/170\n",
            "307/307 [==============================] - 1s 2ms/step - loss: 0.5397 - acc: 0.8678 - val_loss: 0.5588 - val_acc: 0.8607\n",
            "Epoch 114/170\n",
            "307/307 [==============================] - 1s 2ms/step - loss: 0.5339 - acc: 0.8681 - val_loss: 0.5531 - val_acc: 0.8607\n",
            "Epoch 115/170\n",
            "307/307 [==============================] - 1s 2ms/step - loss: 0.5281 - acc: 0.8694 - val_loss: 0.5474 - val_acc: 0.8634\n",
            "Epoch 116/170\n",
            "307/307 [==============================] - 1s 2ms/step - loss: 0.5224 - acc: 0.8703 - val_loss: 0.5417 - val_acc: 0.8634\n",
            "Epoch 117/170\n",
            "307/307 [==============================] - 1s 2ms/step - loss: 0.5168 - acc: 0.8706 - val_loss: 0.5362 - val_acc: 0.8638\n",
            "Epoch 118/170\n",
            "307/307 [==============================] - 1s 2ms/step - loss: 0.5113 - acc: 0.8720 - val_loss: 0.5307 - val_acc: 0.8655\n",
            "Epoch 119/170\n",
            "307/307 [==============================] - 1s 2ms/step - loss: 0.5059 - acc: 0.8719 - val_loss: 0.5253 - val_acc: 0.8673\n",
            "Epoch 120/170\n",
            "307/307 [==============================] - 1s 2ms/step - loss: 0.5005 - acc: 0.8757 - val_loss: 0.5200 - val_acc: 0.8729\n",
            "Epoch 121/170\n",
            "307/307 [==============================] - 1s 2ms/step - loss: 0.4952 - acc: 0.8782 - val_loss: 0.5147 - val_acc: 0.8747\n",
            "Epoch 122/170\n",
            "307/307 [==============================] - 1s 2ms/step - loss: 0.4899 - acc: 0.8792 - val_loss: 0.5095 - val_acc: 0.8760\n",
            "Epoch 123/170\n",
            "307/307 [==============================] - 1s 2ms/step - loss: 0.4847 - acc: 0.8823 - val_loss: 0.5043 - val_acc: 0.8768\n",
            "Epoch 124/170\n",
            "307/307 [==============================] - 1s 2ms/step - loss: 0.4796 - acc: 0.8848 - val_loss: 0.4992 - val_acc: 0.8790\n",
            "Epoch 125/170\n",
            "307/307 [==============================] - 1s 2ms/step - loss: 0.4746 - acc: 0.8875 - val_loss: 0.4942 - val_acc: 0.8873\n",
            "Epoch 126/170\n",
            "307/307 [==============================] - 1s 2ms/step - loss: 0.4696 - acc: 0.8924 - val_loss: 0.4893 - val_acc: 0.8938\n",
            "Epoch 127/170\n",
            "307/307 [==============================] - 1s 2ms/step - loss: 0.4647 - acc: 0.8987 - val_loss: 0.4844 - val_acc: 0.8943\n",
            "Epoch 128/170\n",
            "307/307 [==============================] - 1s 2ms/step - loss: 0.4599 - acc: 0.8987 - val_loss: 0.4796 - val_acc: 0.8943\n",
            "Epoch 129/170\n",
            "307/307 [==============================] - 1s 3ms/step - loss: 0.4551 - acc: 0.8991 - val_loss: 0.4749 - val_acc: 0.8947\n",
            "Epoch 130/170\n",
            "307/307 [==============================] - 1s 3ms/step - loss: 0.4504 - acc: 0.8993 - val_loss: 0.4702 - val_acc: 0.8947\n",
            "Epoch 131/170\n",
            "307/307 [==============================] - 1s 3ms/step - loss: 0.4458 - acc: 0.9010 - val_loss: 0.4656 - val_acc: 0.8977\n",
            "Epoch 132/170\n",
            "307/307 [==============================] - 1s 3ms/step - loss: 0.4412 - acc: 0.9072 - val_loss: 0.4611 - val_acc: 0.9034\n",
            "Epoch 133/170\n",
            "307/307 [==============================] - 1s 3ms/step - loss: 0.4367 - acc: 0.9112 - val_loss: 0.4566 - val_acc: 0.9034\n",
            "Epoch 134/170\n",
            "307/307 [==============================] - 1s 3ms/step - loss: 0.4323 - acc: 0.9115 - val_loss: 0.4522 - val_acc: 0.9034\n",
            "Epoch 135/170\n",
            "307/307 [==============================] - 1s 3ms/step - loss: 0.4279 - acc: 0.9117 - val_loss: 0.4479 - val_acc: 0.9034\n",
            "Epoch 136/170\n",
            "307/307 [==============================] - 1s 3ms/step - loss: 0.4236 - acc: 0.9122 - val_loss: 0.4436 - val_acc: 0.9038\n",
            "Epoch 137/170\n",
            "307/307 [==============================] - 1s 3ms/step - loss: 0.4194 - acc: 0.9129 - val_loss: 0.4395 - val_acc: 0.9056\n",
            "Epoch 138/170\n",
            "307/307 [==============================] - 1s 3ms/step - loss: 0.4153 - acc: 0.9150 - val_loss: 0.4353 - val_acc: 0.9069\n",
            "Epoch 139/170\n",
            "307/307 [==============================] - 1s 3ms/step - loss: 0.4112 - acc: 0.9161 - val_loss: 0.4312 - val_acc: 0.9091\n",
            "Epoch 140/170\n",
            "307/307 [==============================] - 1s 3ms/step - loss: 0.4072 - acc: 0.9177 - val_loss: 0.4272 - val_acc: 0.9095\n",
            "Epoch 141/170\n",
            "307/307 [==============================] - 1s 2ms/step - loss: 0.4032 - acc: 0.9200 - val_loss: 0.4233 - val_acc: 0.9112\n",
            "Epoch 142/170\n",
            "307/307 [==============================] - 1s 2ms/step - loss: 0.3994 - acc: 0.9234 - val_loss: 0.4194 - val_acc: 0.9169\n",
            "Epoch 143/170\n",
            "307/307 [==============================] - 1s 2ms/step - loss: 0.3956 - acc: 0.9245 - val_loss: 0.4156 - val_acc: 0.9191\n",
            "Epoch 144/170\n",
            "307/307 [==============================] - 1s 2ms/step - loss: 0.3918 - acc: 0.9256 - val_loss: 0.4119 - val_acc: 0.9195\n",
            "Epoch 145/170\n",
            "307/307 [==============================] - 1s 2ms/step - loss: 0.3881 - acc: 0.9275 - val_loss: 0.4082 - val_acc: 0.9199\n",
            "Epoch 146/170\n",
            "307/307 [==============================] - 1s 2ms/step - loss: 0.3845 - acc: 0.9276 - val_loss: 0.4046 - val_acc: 0.9199\n",
            "Epoch 147/170\n",
            "307/307 [==============================] - 1s 2ms/step - loss: 0.3810 - acc: 0.9278 - val_loss: 0.4011 - val_acc: 0.9199\n",
            "Epoch 148/170\n",
            "307/307 [==============================] - 1s 2ms/step - loss: 0.3775 - acc: 0.9279 - val_loss: 0.3976 - val_acc: 0.9199\n",
            "Epoch 149/170\n",
            "307/307 [==============================] - 1s 2ms/step - loss: 0.3740 - acc: 0.9279 - val_loss: 0.3942 - val_acc: 0.9204\n",
            "Epoch 150/170\n",
            "307/307 [==============================] - 1s 2ms/step - loss: 0.3707 - acc: 0.9279 - val_loss: 0.3908 - val_acc: 0.9204\n",
            "Epoch 151/170\n",
            "307/307 [==============================] - 1s 2ms/step - loss: 0.3673 - acc: 0.9281 - val_loss: 0.3875 - val_acc: 0.9225\n",
            "Epoch 152/170\n",
            "307/307 [==============================] - 1s 2ms/step - loss: 0.3641 - acc: 0.9293 - val_loss: 0.3843 - val_acc: 0.9225\n",
            "Epoch 153/170\n",
            "307/307 [==============================] - 1s 2ms/step - loss: 0.3608 - acc: 0.9303 - val_loss: 0.3811 - val_acc: 0.9247\n",
            "Epoch 154/170\n",
            "307/307 [==============================] - 1s 2ms/step - loss: 0.3577 - acc: 0.9302 - val_loss: 0.3779 - val_acc: 0.9247\n",
            "Epoch 155/170\n",
            "307/307 [==============================] - 1s 2ms/step - loss: 0.3546 - acc: 0.9306 - val_loss: 0.3748 - val_acc: 0.9247\n",
            "Epoch 156/170\n",
            "307/307 [==============================] - 1s 2ms/step - loss: 0.3515 - acc: 0.9308 - val_loss: 0.3718 - val_acc: 0.9247\n",
            "Epoch 157/170\n",
            "307/307 [==============================] - 1s 2ms/step - loss: 0.3486 - acc: 0.9310 - val_loss: 0.3688 - val_acc: 0.9256\n",
            "Epoch 158/170\n",
            "307/307 [==============================] - 1s 2ms/step - loss: 0.3456 - acc: 0.9336 - val_loss: 0.3659 - val_acc: 0.9299\n",
            "Epoch 159/170\n",
            "307/307 [==============================] - 1s 2ms/step - loss: 0.3427 - acc: 0.9346 - val_loss: 0.3631 - val_acc: 0.9308\n",
            "Epoch 160/170\n",
            "307/307 [==============================] - 1s 2ms/step - loss: 0.3399 - acc: 0.9347 - val_loss: 0.3602 - val_acc: 0.9308\n",
            "Epoch 161/170\n",
            "307/307 [==============================] - 1s 2ms/step - loss: 0.3371 - acc: 0.9347 - val_loss: 0.3574 - val_acc: 0.9308\n",
            "Epoch 162/170\n",
            "307/307 [==============================] - 1s 2ms/step - loss: 0.3343 - acc: 0.9347 - val_loss: 0.3547 - val_acc: 0.9308\n",
            "Epoch 163/170\n",
            "307/307 [==============================] - 1s 2ms/step - loss: 0.3316 - acc: 0.9346 - val_loss: 0.3519 - val_acc: 0.9308\n",
            "Epoch 164/170\n",
            "307/307 [==============================] - 1s 2ms/step - loss: 0.3289 - acc: 0.9354 - val_loss: 0.3493 - val_acc: 0.9312\n",
            "Epoch 165/170\n",
            "307/307 [==============================] - 1s 2ms/step - loss: 0.3263 - acc: 0.9358 - val_loss: 0.3467 - val_acc: 0.9326\n",
            "Epoch 166/170\n",
            "307/307 [==============================] - 1s 2ms/step - loss: 0.3237 - acc: 0.9367 - val_loss: 0.3441 - val_acc: 0.9326\n",
            "Epoch 167/170\n",
            "307/307 [==============================] - 1s 2ms/step - loss: 0.3212 - acc: 0.9364 - val_loss: 0.3415 - val_acc: 0.9326\n",
            "Epoch 168/170\n",
            "307/307 [==============================] - 1s 2ms/step - loss: 0.3187 - acc: 0.9365 - val_loss: 0.3391 - val_acc: 0.9326\n",
            "Epoch 169/170\n",
            "307/307 [==============================] - 1s 2ms/step - loss: 0.3162 - acc: 0.9367 - val_loss: 0.3366 - val_acc: 0.9330\n",
            "Epoch 170/170\n",
            "307/307 [==============================] - 1s 2ms/step - loss: 0.3138 - acc: 0.9381 - val_loss: 0.3342 - val_acc: 0.9343\n",
            "<bound method Model.summary of <tensorflow.python.keras.engine.sequential.Sequential object at 0x7fd8436822e8>>\n",
            "Test Accuracy:\n",
            "3446/3446 [==============================] - 5s 2ms/step - loss: 0.3170 - acc: 0.9338\n",
            "Random Sample Accuracy:\n",
            " 1/50 [..............................] - ETA: 0s - loss: 0.9489 - acc: 0.6333WARNING:tensorflow:Callbacks method `on_test_batch_end` is slow compared to the batch time (batch time: 0.0013s vs `on_test_batch_end` time: 0.0021s). Check your callbacks.\n",
            "50/50 [==============================] - 0s 2ms/step - loss: 0.8247 - acc: 0.6973\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 1080x432 with 2 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "__MpwHll6CmX"
      },
      "source": [
        "Model 6 had the best performance on the random sample, with about 74% accuracy."
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "W-VgBuAfywOp"
      },
      "source": [
        "## Confusion Matrix"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "DEyOaI696aJh"
      },
      "source": [
        "We'll create a confusion matrix to get a fuller sense of how the model performed on the random sample.  "
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "l8EMtKM143FT"
      },
      "source": [
        "label_preds = model4.predict_classes(r_descriptor_onehot)\n",
        "label_preds_onehot = to_categorical(label_preds)"
      ],
      "execution_count": 218,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "fiHQNsVupk2W"
      },
      "source": [
        "cf_matrix = confusion_matrix(r_agency_cat, label_preds, normalize='true')"
      ],
      "execution_count": 220,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 805
        },
        "id": "9VxZFJDWumpA",
        "outputId": "5e1d3b08-c687-4b44-ca86-810be4db07f3"
      },
      "source": [
        "plt.figure(figsize=(13, 13))\n",
        "plt.title('Confusion Matrix: Predictions on Random Sample', \n",
        "          fontdict={'fontsize': 16, 'family': 'serif', 'color': '#58508d', 'weight': 'bold'})\n",
        "\n",
        "sns.heatmap(cf_matrix, annot=True, xticklabels=le.classes_, yticklabels=le.classes_,\n",
        "            fmt='.2%', cmap='Reds')\n",
        "plt.ylabel('True Agency')\n",
        "plt.xlabel('Predicted Agency')"
      ],
      "execution_count": 222,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "Text(0.5, 96.0, 'Predicted Agency')"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 222
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 936x936 with 2 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "urTWxg5jEoAu"
      },
      "source": [
        "The model correctly classified 93.8% of NYPD calls, 80.5% of HBD calls, and 62.7% of DPR calls, but still mis-classifies many calls in the minority classes."
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "lB5-ByaiIYCx"
      },
      "source": [
        "# LSTM Model"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "cx6z6lqAbCpk"
      },
      "source": [
        "Long Short Term Memory (LSTM) recurrent neural network models are appropriate when the order of the input data matters. They are especially useful when working with natural language data, given that the order of words in a grouping of words can significantly change their meaning. Most of the 311 call descriptions are short, so there may be limited variation in the order of words within them. For perspective, the longest descriptor has been printed below."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "CGv2NyIxdIGp",
        "outputId": "cd280be3-37c9-4c43-9a8a-f0c32777939d"
      },
      "source": [
        "#create dictionary and update it with each unique descriptor and its length\n",
        "lengths = {} \n",
        "for descriptor in df.descriptor:\n",
        "  if descriptor not in lengths:\n",
        "    lengths[descriptor] = len(descriptor)\n",
        "\n",
        "#Get descriptor with the highest length value and print it\n",
        "for descriptor, length in lengths.items(): \n",
        "  if len(descriptor) == max(lengths.values()):\n",
        "    print('Descriptor:', descriptor)\n",
        "    print('Length:', length)"
      ],
      "execution_count": 55,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Descriptor: safety netting guard rails  damaged inadequate none  stories  feet or less\n",
            "Length: 74\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "oXOEWvZagy8B"
      },
      "source": [
        "The longest descriptor is only 74 characters. If there is little variation in the order of descriptor words due to their short length, we may not see much improvement in an LSTM model, but possible improvement is nonetheless worth investigating."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "H0WgHIoydZxu"
      },
      "source": [
        "descriptors = X\n",
        "tokenizer = Tokenizer() \n",
        "\n",
        "tokenizer.fit_on_texts(descriptors) \n",
        "descriptor_onehot = tokenizer.texts_to_sequences(descriptors) \n",
        "\n",
        "#padding sequences ensures all sequences have the same length\n",
        "descriptor_onehot = pad_sequences(descriptor_onehot) "
      ],
      "execution_count": 73,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "ULk7tZtHg3KR"
      },
      "source": [
        "#split the test and train data again with new descriptor data\n",
        "test, train, label_test, label_train = train_test_split(descriptor_onehot, \n",
        "                                                        agency_onehot, \n",
        "                                                        test_size=.1, \n",
        "                                                        random_state=6)"
      ],
      "execution_count": 74,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "0O3OwrVGv9JD"
      },
      "source": [
        "#store the shape in an object that will be used in the model's input dimensions\n",
        "train_shape = train.shape"
      ],
      "execution_count": 97,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "6PfhxhxrseIe"
      },
      "source": [
        "#create a model using an LSTM layer, with dropout to prevent overfitting\n",
        "\n",
        "opt = tf.keras.optimizers.SGD(learning_rate=0.01)\n",
        "model = Sequential()\n",
        "model.add(layers.Embedding(input_dim=train_shape[0], output_dim=100))\n",
        "model.add(layers.LSTM(64, dropout=0.1, activation='relu',return_sequences=True))\n",
        "model.add(layers.GRU(14, recurrent_dropout=0.2, dropout=0.2))\n",
        "model.add(Dropout(.2))\n",
        "model.add(layers.Dense(14, activation='softmax'))\n",
        "model.compile(loss='categorical_crossentropy', optimizer=opt, metrics=['accuracy'])"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "5Qy2fvs51KDo"
      },
      "source": [
        "We'll fit the LSTM model with an early stopping callback. This model has a longer training time than those built in the previous section. Implementing early stopping will ensure that the model does not continue to train after its improvement with each epoch becomes insignificant."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "-kRHhCxQvRZp",
        "outputId": "38c0ad6c-c74a-45f6-fca2-8437da60e700"
      },
      "source": [
        "early_stopping = EarlyStopping(monitor='val_loss', patience=3)\n",
        "history = model.fit(train, label_train, validation_data=(test, label_test), \n",
        "                    epochs=150, batch_size=30, callbacks=[early_stopping])"
      ],
      "execution_count": 168,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Epoch 1/150\n",
            "383/383 [==============================] - 42s 110ms/step - loss: 2.0293 - accuracy: 0.5166 - val_loss: 1.7932 - val_accuracy: 0.5231\n",
            "Epoch 2/150\n",
            "383/383 [==============================] - 41s 108ms/step - loss: 1.7864 - accuracy: 0.5194 - val_loss: 1.7156 - val_accuracy: 0.5231\n",
            "Epoch 3/150\n",
            "383/383 [==============================] - 42s 109ms/step - loss: 1.7281 - accuracy: 0.5194 - val_loss: 1.6680 - val_accuracy: 0.5231\n",
            "Epoch 4/150\n",
            "383/383 [==============================] - 43s 111ms/step - loss: 1.6803 - accuracy: 0.5194 - val_loss: 1.6185 - val_accuracy: 0.5231\n",
            "Epoch 5/150\n",
            "383/383 [==============================] - 42s 110ms/step - loss: 1.6163 - accuracy: 0.5195 - val_loss: 1.5367 - val_accuracy: 0.5231\n",
            "Epoch 6/150\n",
            "383/383 [==============================] - 42s 108ms/step - loss: 1.5202 - accuracy: 0.5258 - val_loss: 1.4179 - val_accuracy: 0.5393\n",
            "Epoch 7/150\n",
            "383/383 [==============================] - 42s 109ms/step - loss: 1.3972 - accuracy: 0.5586 - val_loss: 1.3041 - val_accuracy: 0.5758\n",
            "Epoch 8/150\n",
            "383/383 [==============================] - 42s 109ms/step - loss: 1.2828 - accuracy: 0.5863 - val_loss: 1.1705 - val_accuracy: 0.5949\n",
            "Epoch 9/150\n",
            "383/383 [==============================] - 42s 108ms/step - loss: 1.1802 - accuracy: 0.6242 - val_loss: 1.0712 - val_accuracy: 0.6833\n",
            "Epoch 10/150\n",
            "383/383 [==============================] - 42s 109ms/step - loss: 1.0980 - accuracy: 0.6510 - val_loss: 1.0009 - val_accuracy: 0.6881\n",
            "Epoch 11/150\n",
            "383/383 [==============================] - 43s 112ms/step - loss: 1.0345 - accuracy: 0.6724 - val_loss: 0.9622 - val_accuracy: 0.6923\n",
            "Epoch 12/150\n",
            "383/383 [==============================] - 42s 110ms/step - loss: 0.9902 - accuracy: 0.6850 - val_loss: 0.9195 - val_accuracy: 0.6973\n",
            "Epoch 13/150\n",
            "383/383 [==============================] - 45s 117ms/step - loss: 0.9545 - accuracy: 0.6909 - val_loss: 0.8780 - val_accuracy: 0.7037\n",
            "Epoch 14/150\n",
            "383/383 [==============================] - 43s 111ms/step - loss: 0.9261 - accuracy: 0.6951 - val_loss: 0.8498 - val_accuracy: 0.7033\n",
            "Epoch 15/150\n",
            "383/383 [==============================] - 42s 111ms/step - loss: 0.9009 - accuracy: 0.7008 - val_loss: 0.8239 - val_accuracy: 0.7107\n",
            "Epoch 16/150\n",
            "383/383 [==============================] - 42s 108ms/step - loss: 0.8777 - accuracy: 0.7108 - val_loss: 0.8011 - val_accuracy: 0.7222\n",
            "Epoch 17/150\n",
            "383/383 [==============================] - 41s 108ms/step - loss: 0.8529 - accuracy: 0.7190 - val_loss: 0.7712 - val_accuracy: 0.7324\n",
            "Epoch 18/150\n",
            "383/383 [==============================] - 42s 110ms/step - loss: 0.8269 - accuracy: 0.7267 - val_loss: 0.7472 - val_accuracy: 0.7477\n",
            "Epoch 19/150\n",
            "383/383 [==============================] - 41s 108ms/step - loss: 0.8035 - accuracy: 0.7359 - val_loss: 0.7171 - val_accuracy: 0.7625\n",
            "Epoch 20/150\n",
            "383/383 [==============================] - 41s 107ms/step - loss: 0.7756 - accuracy: 0.7464 - val_loss: 0.6902 - val_accuracy: 0.7750\n",
            "Epoch 21/150\n",
            "383/383 [==============================] - 41s 107ms/step - loss: 0.7508 - accuracy: 0.7587 - val_loss: 0.6611 - val_accuracy: 0.7764\n",
            "Epoch 22/150\n",
            "383/383 [==============================] - 41s 108ms/step - loss: 0.7095 - accuracy: 0.7753 - val_loss: 0.6096 - val_accuracy: 0.7970\n",
            "Epoch 23/150\n",
            "383/383 [==============================] - 41s 107ms/step - loss: 0.6841 - accuracy: 0.7862 - val_loss: 0.5760 - val_accuracy: 0.8437\n",
            "Epoch 24/150\n",
            "383/383 [==============================] - 41s 106ms/step - loss: 0.6597 - accuracy: 0.7922 - val_loss: 0.5464 - val_accuracy: 0.8482\n",
            "Epoch 25/150\n",
            "383/383 [==============================] - 41s 106ms/step - loss: 0.6229 - accuracy: 0.8106 - val_loss: 0.5189 - val_accuracy: 0.8558\n",
            "Epoch 26/150\n",
            "383/383 [==============================] - 41s 108ms/step - loss: 0.5942 - accuracy: 0.8230 - val_loss: 0.4918 - val_accuracy: 0.8659\n",
            "Epoch 27/150\n",
            "383/383 [==============================] - 41s 106ms/step - loss: 0.5746 - accuracy: 0.8296 - val_loss: 0.4646 - val_accuracy: 0.8723\n",
            "Epoch 28/150\n",
            "383/383 [==============================] - 41s 106ms/step - loss: 0.5514 - accuracy: 0.8338 - val_loss: 0.4624 - val_accuracy: 0.8773\n",
            "Epoch 29/150\n",
            "383/383 [==============================] - 41s 107ms/step - loss: 0.5267 - accuracy: 0.8419 - val_loss: 0.4278 - val_accuracy: 0.8781\n",
            "Epoch 30/150\n",
            "383/383 [==============================] - 41s 107ms/step - loss: 0.5022 - accuracy: 0.8530 - val_loss: 0.4069 - val_accuracy: 0.8816\n",
            "Epoch 31/150\n",
            "383/383 [==============================] - 41s 106ms/step - loss: 0.4888 - accuracy: 0.8580 - val_loss: 0.4042 - val_accuracy: 0.8886\n",
            "Epoch 32/150\n",
            "383/383 [==============================] - 40s 106ms/step - loss: 0.4739 - accuracy: 0.8590 - val_loss: 0.3754 - val_accuracy: 0.8952\n",
            "Epoch 33/150\n",
            "383/383 [==============================] - 41s 108ms/step - loss: 0.4552 - accuracy: 0.8633 - val_loss: 0.3585 - val_accuracy: 0.8943\n",
            "Epoch 34/150\n",
            "383/383 [==============================] - 41s 107ms/step - loss: 0.4443 - accuracy: 0.8709 - val_loss: 0.3621 - val_accuracy: 0.9050\n",
            "Epoch 35/150\n",
            "383/383 [==============================] - 40s 106ms/step - loss: 0.4265 - accuracy: 0.8768 - val_loss: 0.3424 - val_accuracy: 0.8984\n",
            "Epoch 36/150\n",
            "383/383 [==============================] - 41s 106ms/step - loss: 0.4167 - accuracy: 0.8812 - val_loss: 0.3298 - val_accuracy: 0.9194\n",
            "Epoch 37/150\n",
            "383/383 [==============================] - 41s 107ms/step - loss: 0.3957 - accuracy: 0.8884 - val_loss: 0.3105 - val_accuracy: 0.9191\n",
            "Epoch 38/150\n",
            "383/383 [==============================] - 41s 106ms/step - loss: 0.3830 - accuracy: 0.8953 - val_loss: 0.2971 - val_accuracy: 0.9303\n",
            "Epoch 39/150\n",
            "383/383 [==============================] - 40s 105ms/step - loss: 0.3745 - accuracy: 0.8993 - val_loss: 0.3127 - val_accuracy: 0.9165\n",
            "Epoch 40/150\n",
            "383/383 [==============================] - 41s 107ms/step - loss: 0.3613 - accuracy: 0.9042 - val_loss: 0.2958 - val_accuracy: 0.9253\n",
            "Epoch 41/150\n",
            "383/383 [==============================] - 41s 108ms/step - loss: 0.3540 - accuracy: 0.9059 - val_loss: 0.2601 - val_accuracy: 0.9471\n",
            "Epoch 42/150\n",
            "383/383 [==============================] - 41s 108ms/step - loss: 0.3363 - accuracy: 0.9129 - val_loss: 0.2464 - val_accuracy: 0.9471\n",
            "Epoch 43/150\n",
            "383/383 [==============================] - 42s 109ms/step - loss: 0.3327 - accuracy: 0.9135 - val_loss: 0.2392 - val_accuracy: 0.9413\n",
            "Epoch 44/150\n",
            "383/383 [==============================] - 44s 114ms/step - loss: 0.3205 - accuracy: 0.9189 - val_loss: 0.2459 - val_accuracy: 0.9539\n",
            "Epoch 45/150\n",
            "383/383 [==============================] - 41s 108ms/step - loss: 0.3084 - accuracy: 0.9212 - val_loss: 0.2231 - val_accuracy: 0.9479\n",
            "Epoch 46/150\n",
            "383/383 [==============================] - 41s 108ms/step - loss: 0.3025 - accuracy: 0.9232 - val_loss: 0.2310 - val_accuracy: 0.9336\n",
            "Epoch 47/150\n",
            "383/383 [==============================] - 41s 108ms/step - loss: 0.2943 - accuracy: 0.9242 - val_loss: 0.2099 - val_accuracy: 0.9543\n",
            "Epoch 48/150\n",
            "383/383 [==============================] - 42s 109ms/step - loss: 0.2818 - accuracy: 0.9277 - val_loss: 0.2110 - val_accuracy: 0.9477\n",
            "Epoch 49/150\n",
            "383/383 [==============================] - 41s 107ms/step - loss: 0.2727 - accuracy: 0.9304 - val_loss: 0.1956 - val_accuracy: 0.9552\n",
            "Epoch 50/150\n",
            "383/383 [==============================] - 41s 107ms/step - loss: 0.2709 - accuracy: 0.9321 - val_loss: 0.2056 - val_accuracy: 0.9470\n",
            "Epoch 51/150\n",
            "383/383 [==============================] - 41s 107ms/step - loss: 0.2638 - accuracy: 0.9353 - val_loss: 0.1891 - val_accuracy: 0.9602\n",
            "Epoch 52/150\n",
            "383/383 [==============================] - 41s 108ms/step - loss: 0.2584 - accuracy: 0.9354 - val_loss: 0.1794 - val_accuracy: 0.9604\n",
            "Epoch 53/150\n",
            "383/383 [==============================] - 41s 107ms/step - loss: 0.2531 - accuracy: 0.9357 - val_loss: 0.1767 - val_accuracy: 0.9631\n",
            "Epoch 54/150\n",
            "383/383 [==============================] - 41s 106ms/step - loss: 0.2508 - accuracy: 0.9372 - val_loss: 0.1735 - val_accuracy: 0.9621\n",
            "Epoch 55/150\n",
            "383/383 [==============================] - 41s 106ms/step - loss: 0.2449 - accuracy: 0.9392 - val_loss: 0.1663 - val_accuracy: 0.9619\n",
            "Epoch 56/150\n",
            "383/383 [==============================] - 41s 108ms/step - loss: 0.2329 - accuracy: 0.9429 - val_loss: 0.1612 - val_accuracy: 0.9633\n",
            "Epoch 57/150\n",
            "383/383 [==============================] - 41s 106ms/step - loss: 0.2273 - accuracy: 0.9422 - val_loss: 0.1599 - val_accuracy: 0.9642\n",
            "Epoch 58/150\n",
            "383/383 [==============================] - 41s 106ms/step - loss: 0.2263 - accuracy: 0.9436 - val_loss: 0.1586 - val_accuracy: 0.9633\n",
            "Epoch 59/150\n",
            "383/383 [==============================] - 41s 106ms/step - loss: 0.2221 - accuracy: 0.9436 - val_loss: 0.1476 - val_accuracy: 0.9668\n",
            "Epoch 60/150\n",
            "383/383 [==============================] - 41s 106ms/step - loss: 0.2205 - accuracy: 0.9460 - val_loss: 0.1542 - val_accuracy: 0.9645\n",
            "Epoch 61/150\n",
            "383/383 [==============================] - 41s 106ms/step - loss: 0.2144 - accuracy: 0.9448 - val_loss: 0.1509 - val_accuracy: 0.9644\n",
            "Epoch 62/150\n",
            "383/383 [==============================] - 41s 107ms/step - loss: 0.2099 - accuracy: 0.9485 - val_loss: 0.1443 - val_accuracy: 0.9688\n",
            "Epoch 63/150\n",
            "383/383 [==============================] - 41s 106ms/step - loss: 0.2032 - accuracy: 0.9478 - val_loss: 0.1412 - val_accuracy: 0.9661\n",
            "Epoch 64/150\n",
            "383/383 [==============================] - 41s 108ms/step - loss: 0.1994 - accuracy: 0.9495 - val_loss: 0.1353 - val_accuracy: 0.9653\n",
            "Epoch 65/150\n",
            "383/383 [==============================] - 41s 106ms/step - loss: 0.2011 - accuracy: 0.9480 - val_loss: 0.1347 - val_accuracy: 0.9660\n",
            "Epoch 66/150\n",
            "383/383 [==============================] - 41s 107ms/step - loss: 0.1916 - accuracy: 0.9520 - val_loss: 0.1285 - val_accuracy: 0.9671\n",
            "Epoch 67/150\n",
            "383/383 [==============================] - 41s 106ms/step - loss: 0.1871 - accuracy: 0.9525 - val_loss: 0.1261 - val_accuracy: 0.9675\n",
            "Epoch 68/150\n",
            "383/383 [==============================] - 41s 107ms/step - loss: 0.1883 - accuracy: 0.9508 - val_loss: 0.1383 - val_accuracy: 0.9641\n",
            "Epoch 69/150\n",
            "383/383 [==============================] - 41s 107ms/step - loss: 0.1864 - accuracy: 0.9510 - val_loss: 0.1204 - val_accuracy: 0.9689\n",
            "Epoch 70/150\n",
            "383/383 [==============================] - 41s 106ms/step - loss: 0.1823 - accuracy: 0.9546 - val_loss: 0.1248 - val_accuracy: 0.9730\n",
            "Epoch 71/150\n",
            "383/383 [==============================] - 42s 109ms/step - loss: 0.1785 - accuracy: 0.9533 - val_loss: 0.1213 - val_accuracy: 0.9690\n",
            "Epoch 72/150\n",
            "383/383 [==============================] - 41s 106ms/step - loss: 0.1702 - accuracy: 0.9569 - val_loss: 0.1177 - val_accuracy: 0.9741\n",
            "Epoch 73/150\n",
            "383/383 [==============================] - 41s 107ms/step - loss: 0.1753 - accuracy: 0.9540 - val_loss: 0.1145 - val_accuracy: 0.9701\n",
            "Epoch 74/150\n",
            "383/383 [==============================] - 41s 107ms/step - loss: 0.1702 - accuracy: 0.9557 - val_loss: 0.1267 - val_accuracy: 0.9673\n",
            "Epoch 75/150\n",
            "383/383 [==============================] - 40s 106ms/step - loss: 0.1650 - accuracy: 0.9560 - val_loss: 0.1095 - val_accuracy: 0.9758\n",
            "Epoch 76/150\n",
            "383/383 [==============================] - 41s 106ms/step - loss: 0.1641 - accuracy: 0.9573 - val_loss: 0.1099 - val_accuracy: 0.9749\n",
            "Epoch 77/150\n",
            "383/383 [==============================] - 40s 105ms/step - loss: 0.1629 - accuracy: 0.9594 - val_loss: 0.1072 - val_accuracy: 0.9721\n",
            "Epoch 78/150\n",
            "383/383 [==============================] - 41s 106ms/step - loss: 0.1608 - accuracy: 0.9586 - val_loss: 0.1177 - val_accuracy: 0.9692\n",
            "Epoch 79/150\n",
            "383/383 [==============================] - 41s 108ms/step - loss: 0.1539 - accuracy: 0.9606 - val_loss: 0.1068 - val_accuracy: 0.9749\n",
            "Epoch 80/150\n",
            "383/383 [==============================] - 41s 106ms/step - loss: 0.1537 - accuracy: 0.9604 - val_loss: 0.1038 - val_accuracy: 0.9767\n",
            "Epoch 81/150\n",
            "383/383 [==============================] - 41s 107ms/step - loss: 0.1510 - accuracy: 0.9600 - val_loss: 0.1051 - val_accuracy: 0.9737\n",
            "Epoch 82/150\n",
            "383/383 [==============================] - 41s 106ms/step - loss: 0.1485 - accuracy: 0.9620 - val_loss: 0.0997 - val_accuracy: 0.9785\n",
            "Epoch 83/150\n",
            "383/383 [==============================] - 41s 106ms/step - loss: 0.1462 - accuracy: 0.9618 - val_loss: 0.1063 - val_accuracy: 0.9745\n",
            "Epoch 84/150\n",
            "383/383 [==============================] - 40s 105ms/step - loss: 0.1454 - accuracy: 0.9639 - val_loss: 0.0958 - val_accuracy: 0.9785\n",
            "Epoch 85/150\n",
            "383/383 [==============================] - 40s 105ms/step - loss: 0.1436 - accuracy: 0.9637 - val_loss: 0.0964 - val_accuracy: 0.9781\n",
            "Epoch 86/150\n",
            "383/383 [==============================] - 41s 108ms/step - loss: 0.1412 - accuracy: 0.9621 - val_loss: 0.1034 - val_accuracy: 0.9763\n",
            "Epoch 87/150\n",
            "383/383 [==============================] - 41s 106ms/step - loss: 0.1428 - accuracy: 0.9614 - val_loss: 0.0951 - val_accuracy: 0.9785\n",
            "Epoch 88/150\n",
            "383/383 [==============================] - 40s 105ms/step - loss: 0.1333 - accuracy: 0.9650 - val_loss: 0.0921 - val_accuracy: 0.9788\n",
            "Epoch 89/150\n",
            "383/383 [==============================] - 41s 107ms/step - loss: 0.1356 - accuracy: 0.9653 - val_loss: 0.0957 - val_accuracy: 0.9760\n",
            "Epoch 90/150\n",
            "383/383 [==============================] - 40s 106ms/step - loss: 0.1328 - accuracy: 0.9660 - val_loss: 0.0942 - val_accuracy: 0.9786\n",
            "Epoch 91/150\n",
            "383/383 [==============================] - 41s 106ms/step - loss: 0.1333 - accuracy: 0.9657 - val_loss: 0.0971 - val_accuracy: 0.9782\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "eZLVzrZW2yco"
      },
      "source": [
        "With the early stopping callback, the model stopped training after 91 epochs. Below, we use modified code from the `plot_acc_loss()` function to visualize the model's learning history."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 407
        },
        "id": "HTN75iuepZDb",
        "outputId": "7f10d5bc-4d7f-430a-997e-d8b01399c2b4"
      },
      "source": [
        "fig, (ax1, ax2) = plt.subplots(nrows=1,\n",
        "                               ncols=2,\n",
        "                               figsize=(15, 6))\n",
        "\n",
        "ax_font = {'fontsize': 12, 'family': 'serif', \n",
        "             'color': '#58508d', 'weight': 'bold'}\n",
        "title_font = {'fontsize': 14, 'family': 'serif',\n",
        "                'color': '#58508d', 'weight': 'bold'}\n",
        "\n",
        "\n",
        "ax1.plot(history.history['accuracy'], color='#D8A7B1', linewidth=3)\n",
        "ax1.grid(color='g', linestyle='-', linewidth=.09)\n",
        "ax1.plot(history.history['val_accuracy'], color='#05234C', linewidth=3)\n",
        "ax1.set_title('Model Accuracy', fontdict=title_font)\n",
        "ax1.set_ylabel('Accuracy', fontdict=ax_font)\n",
        "ax1.set_xlabel('Number of epochs', fontdict=ax_font)\n",
        "ax1.legend(['train', 'validation'], loc='upper left')\n",
        "\n",
        "\n",
        "ax2.plot(history.history['loss'], color='#D8A7B1', linewidth=3)\n",
        "ax2.grid(color='g', linestyle='-', linewidth=.09)\n",
        "ax2.plot(history.history['val_loss'], color='#05234C', linewidth=3)\n",
        "ax2.set_title('Model Loss', fontdict=title_font)\n",
        "ax2.set_ylabel('Loss', fontdict=ax_font)\n",
        "ax2.set_xlabel('Number of epochs', fontdict=ax_font)\n",
        "ax2.legend(['train', 'validation'], loc='upper right')\n",
        "\n",
        "fig.show()"
      ],
      "execution_count": 174,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 1080x432 with 2 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "6xPrV2ul3dH9"
      },
      "source": [
        "The model's performance on the validation data improved steadily alongside the training data across the 91 epochs. We'll now evaluate its accuracy on the random subset. "
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "sUtwxrgbrKuT"
      },
      "source": [
        "#create random sample test data with padded sequences \n",
        "tokenizer.fit_on_texts(r_descriptors) \n",
        "LSTM_r_descriptor_onehot = tokenizer.texts_to_sequences(r_descriptors)\n",
        "LSTM_r_descriptor_onehot = pad_sequences(LSTM_r_descriptor_onehot)"
      ],
      "execution_count": 183,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "4Kpfl3z6w6tU",
        "outputId": "aa7320df-ac13-4ebc-9bba-0471f4a43cb4"
      },
      "source": [
        "#evaluate accuracy\n",
        "LSTM_random_results = model.evaluate(LSTM_r_descriptor_onehot, r_agency_onehot, batch_size=30)"
      ],
      "execution_count": 184,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "3863/3863 [==============================] - 25s 6ms/step - loss: 1.6404 - accuracy: 0.7057\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "teWJLk-I4DF_"
      },
      "source": [
        "In terms of accuracy scores on the random sample, the performance of the fourth model still exceeds that of the LSTM model. Different parameters could be tested on the LSTM model, but this process would be costly given the substantial increase in training time. "
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "nQCkQBCrLMhL"
      },
      "source": [
        "# Interpreting the Results\n",
        "With 92% accuracy on the resampled test data, and 72.8% accuracy on the random subset, the fourth sequential model had the best overall performance. \n",
        "\n",
        "- The model correctly predicted that roughly half (49%) of calls received a response from the NYPD. \n",
        "- About 7% of predictions were correct HPD classifications.\n",
        "- About 6% were correct DPR classifications. \n",
        "- About 3% were correct DOT classifications, and another 2% were correct DSNY classifications.\n",
        "- Despite being trained on resampled data, the model still mis-classified the responding agency of some calls as the NYPD, HPD, or DOT. About 3% of calls, for example, were HPD but classified as DOT.\n",
        "\n",
        "# Conclusion and Recommendations\n",
        "- Similar classification models can be developed to connect individuals with non-emergency government services by directing them to the appropriate responding agency. This potential application would require training on a larger, more varied set of descriptors. The evaluation of future models may also consider metrics other than accuracy to achieve a fuller understanding of their performance.\n",
        "- Agencies should be attentive to how call volume tends to change based on certain temporal, geographic, and environmental factors. Many of these changes are intuitively expected: widespread tree damage following major weather events, noise calls peaking in the middle of the night, and overall call volume remaining consistently high in the most densely populated borough, Manhattan. \n",
        "- Due to the novelty of COVID-19 regulations, complaints related to COVID-19 have appeared only recently, so particular attention should be directed toward the circumstances surrounding any future peaks.\n",
        "- Given that the majority of non-emergency requests are responded to by the same agency responsible for emergency requests, local stakeholders may wish to evaluate whether the current division of labor in handling 311 calls is optimal for meeting the needs of city residents. Amid growing concerns that law enforcement officers are over-utilized for intervention in non-emergency situations, this could be a fruitful area for further inquiry.\n"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "buv6obDyjy8g"
      },
      "source": [
        "# References \n",
        "[Hyperparameter Optimization with Keras]('https://towardsdatascience.com/hyperparameter-optimization-with-keras-b82e6364ca53')"
      ]
    }
  ]
}