0.0.10 release

kviz/visualizer.py

@@ -27,6 +27,14 @@
 import tensorflow.keras as keras
 
 
+COLORS = np.array(['purple', 'blue'])
+tuples = list(zip(
+    map(plt.Normalize(0, 1), [0, .5, 1]),
+    [COLORS[0], 'white', COLORS[1]])
+)
+CMAP = LinearSegmentedColormap.from_list("", tuples, 100)
+
+
 class Visualizer():
     """
         Class for creating and rendering visualization of Keras
@@ -156,17 +164,8 @@ def _get_int_models(self):
 
         for l in range(len(self.model.layers)):
             int_model = keras.Sequential()
-            inp = self.model.layers[0]._input_shape_arg[0]
-
             for prev_layer in range(l + 1):
-                int_layer = self.model.layers[prev_layer]
-                int_model.add(keras.layers.Dense(
-                    int_layer.units,
-                    input_dim=inp,
-                    activation=int_layer.activation)
-                )
-                int_model.layers[prev_layer].set_weights(int_layer.get_weights())
-                inp = int_layer.units
+                int_model.add(self.model.layers[prev_layer])
             int_model.compile(loss=self.model.loss)
             intermediate_models.append(int_model)
 
@@ -228,14 +227,10 @@ def _snap_decision_boundary(self, X, Y, filename):
                              np.arange(y_min, y_max, h))
         meshData = np.c_[xx.ravel(), yy.ravel()]
 
-        # TODO catch max number of classes
-        colors = np.array([x for x in 'bgrcmyk'])
-
         fig, ax = plt.subplots(frameon=False)
-        ax.scatter(X[:, 0], X[:, 1], color=colors[Y].tolist(), s=100, alpha=.9)
-        Z = self.model.predict(meshData)
-        Z = np.array([int(round(z[0])) for z in Z]).reshape(xx.shape)
-        ax.contourf(xx, yy, Z, alpha=.5, cmap=plt.cm.Paired)
+        ax.scatter(X[:, 0], X[:, 1], color=COLORS[Y].tolist(), s=100, alpha=.9)
+        Z = self.model.predict(meshData).reshape(xx.shape)
+        ax.contourf(xx, yy, Z, alpha=.4, cmap=CMAP)
         fig.savefig(filename + '.png')
         plt.close()
 
@@ -342,7 +337,7 @@ def fit(self, X, Y, snap_freq=10, filename='decision_boundary', duration=1000, *
 
         for _ in range(int(epochs / snap_freq)):
             self.model.fit(X, Y, epochs=snap_freq, **kwargs)
-            # self._int_models = self._get_int_models() TODO: make this function more efficient
+            self._int_models = self._get_int_models()  # TODO: make this function more efficient
             images.append(im.fromarray(self._snap_decision_boundary(X, Y, filename)))
 
         self._convert_gif(images, filename, duration)
@@ -369,12 +364,6 @@ def view_activations_for(self, X, filename='activations', duration=1000, x_color
             None
         """
 
-        cvals = [0, .5, 1]
-        colors = ['purple', 'white', 'blue']
-        norm = plt.Normalize(min(cvals), max(cvals))
-        tuples = list(zip(map(norm, cvals), colors))
-        CMAP = LinearSegmentedColormap.from_list("", tuples, 100)
-
         cvals = [0, 1]
         colors = ['white', 'green']
         norm = plt.Normalize(min(cvals), max(cvals))

setup.py

@@ -12,7 +12,7 @@
 setup(
   name='kviz',
   packages=['kviz'],
-  version='0.0.9',
+  version='0.0.10',
   description='A Library for visualizing keras neural networks',
   install_requires=[
     'tensorflow',

tests/test_visualizer.py

@@ -1,27 +1,25 @@
 import numpy as np
 from tensorflow import keras
 from tensorflow.keras import layers
-import matplotlib.pyplot as plt
 import sklearn.datasets as datasets
 
 from kviz.visualizer import Visualizer
 
 
-def test_dense_no_input():
+def test_render():
     ACTIVATION = "sigmoid"
     model = keras.models.Sequential()
     model.add(layers.Dense(2, input_dim=2))
     model.add(layers.Dense(1, activation=ACTIVATION))
     model.compile(loss="binary_crossentropy")
 
     dg = Visualizer(model)
-    dg.render(filename='test_no_input')
+    dg.render(filename='test_render')
     g1 = dg.get_graph()
     dg.set_graph(g1)
-    dg.get_graph()
 
 
-def test_dense_input_xor():
+def test_view_activations_for():
     ACTIVATION = "sigmoid"
     model = keras.models.Sequential()
     model.add(layers.Dense(3, input_dim=2, activation=ACTIVATION))
@@ -35,74 +33,12 @@ def test_dense_input_xor():
         [1, 1]])
     Y = np.array([x[0] ^ x[1] for x in X])
 
-    model.fit(X, Y, batch_size=4, epochs=1000)
-
-    colors = np.array(['b', 'g'])
-    fig, ax = plt.subplots()
-    ax.scatter(X[:, 0], X[:, 1], color=colors[Y].tolist(), s=50, alpha=0.8)
-
-    h = .02
-    x_min, x_max = X[:, 0].min() - .5, X[:, 0].max() + .5
-    y_min, y_max = X[:, 1].min() - .5, X[:, 1].max() + .5
-    xx, yy = np.meshgrid(np.arange(x_min, x_max, h),
-                         np.arange(y_min, y_max, h))
-    meshData = np.c_[xx.ravel(), yy.ravel()]
-
-    Z = model.predict(meshData)
-    Z = np.array([0 if x < .5 else 1 for x in Z])
-    Z = Z.reshape(xx.shape)
-    ax.contourf(xx, yy, Z, alpha=.3, cmap=plt.cm.Paired)
-    ax.axis('off')
-
-    fig.savefig("test_model_xor.png")
-
-    dg = Visualizer(model)
-    dg.view_activations_for(X, filename='test_input_xor')
-
-
-def test_dense_input_line():
-    ACTIVATION = "sigmoid"
-    model = keras.models.Sequential()
-    model.add(layers.Dense(3, input_dim=2, activation=ACTIVATION))
-    model.add(layers.Dense(1, activation=ACTIVATION))
-    model.compile(loss="binary_crossentropy")
-
-    t, _ = datasets.make_blobs(n_samples=50, centers=[[.5, .5]], cluster_std=.1, random_state=1)
-    X = np.array(t)
-    Y = np.array([1 if x[0] - x[1] >= 0 else 0 for x in X])
-
-    model.fit(X, Y, batch_size=50, epochs=100)
-
-    # see which nodes activate for a given class
-    X0 = X[X[:, 0] - X[:, 1] <= 0]
-    X1 = X[X[:, 0] - X[:, 1] >= 0]
-
-    X = np.concatenate((X0, X1), axis=0)
-
-    fig, ax = plt.subplots()
-    ax.scatter(X0[:, 0], X0[:, 1], color='b', s=10, alpha=0.8)
-    ax.scatter(X1[:, 0], X1[:, 1], facecolors='none', edgecolors='black')
-
-    h = .01
-    x_min, x_max = 0, 1
-    y_min, y_max = 0, 1
-    xx, yy = np.meshgrid(np.arange(x_min, x_max, h),
-                         np.arange(y_min, y_max, h))
-    meshData = np.c_[xx.ravel(), yy.ravel()]
-
-    Z = model.predict(meshData)
-    Z = np.array([0 if x < .5 else 1 for x in Z])
-    Z = Z.reshape(xx.shape)
-    ax.contourf(xx, yy, Z, alpha=.3, cmap=plt.cm.Paired)
-    ax.axis('off')
-
-    fig.savefig("test_model_line.png")
-
     dg = Visualizer(model)
-    dg.view_activations_for(X, filename='test_input_line', duration=300)
+    dg.fit(X, Y, 10, 'test_view_activations_for_decision_boundary', 100, epochs=2000)
+    dg.view_activations_for(X, filename='test_view_activations_for')
 
 
-def test_animate_learning():
+def test_fit():
     ACTIVATION = "sigmoid"
 
     def custom_activation(x):
@@ -118,4 +54,5 @@ def custom_activation(x):
     Y = np.array([1 if x[0]**2 + x[1]**2 >= 1 else 0 for x in X])
 
     dg = Visualizer(model)
-    dg.fit(X, Y, filename='test_animate', verbose=0, batch_size=50)
+    dg.fit(X, Y, 10, 'test_fit', 100, epochs=200, verbose=0, batch_size=50)
+    dg.view_activations_for(X, 'test_fit_activations', 100)