Param

CHANGELOG.md

@@ -3,6 +3,14 @@ All notable changes to this project will be documented in this file.
 
 The format is based on [Keep a Changelog](http://keepachangelog.com/en/1.0.0/).
 
+## [1.1.5] - 2018-07-21
+### Added
+- setting parameters for different models
+
+## [1.1.4] - 2018-07-21
+### Added
+- sample method for LinearRegression model
+
 ## [1.1.3] - 2018-05-25
 ### Fixed
 - HLR fit method sets shared vars if no minibatch_size given

docs/api/modules.rst

@@ -1,7 +1,7 @@
-api
+pymc3_models
 ============
 
 .. toctree::
    :maxdepth: 4
 
-   pymc3_models.models
+   pymc3_models

docs/api/pymc3_models.models.rst

@@ -1,16 +1,19 @@
-models
-=============================
+pymc3\_models.models package
+============================
 
-pymc3\_models\.models\.HierarchicalLogisticRegression module
-------------------------------------------------------------
+Submodules
+----------
+
+pymc3\_models.models.HierarchicalLogisticRegression module
+----------------------------------------------------------
 
 .. automodule:: pymc3_models.models.HierarchicalLogisticRegression
     :members:
     :undoc-members:
     :show-inheritance:
 
-pymc3\_models\.models\.LinearRegression module
-----------------------------------------------
+pymc3\_models.models.LinearRegression module
+--------------------------------------------
 
 .. automodule:: pymc3_models.models.LinearRegression
     :members:

docs/api/pymc3_models.rst

@@ -0,0 +1,29 @@
+pymc3\_models package
+=====================
+
+Subpackages
+-----------
+
+.. toctree::
+
+    pymc3_models.models
+
+Submodules
+----------
+
+pymc3\_models.exc module
+------------------------
+
+.. automodule:: pymc3_models.exc
+    :members:
+    :undoc-members:
+    :show-inheritance:
+
+
+Module contents
+---------------
+
+.. automodule:: pymc3_models
+    :members:
+    :undoc-members:
+    :show-inheritance:

pymc3_models/models/HierarchicalLogisticRegression.py

@@ -13,10 +13,23 @@ class HierarchicalLogisticRegression(BayesianModel):
     Custom Hierachical Logistic Regression built using PyMC3.
     """
 
-    def __init__(self):
+    def __init__(self,
+                 mu_alpha_mu    = 0.0,
+                 mu_alpha_sd    = 100.0,
+                 sigma_alpha_sd = 100.0,
+                 mu_beta_mu     = 0.0,
+                 mu_beta_sd     = 100.0,
+                 sigma_beta_sd  = 100.0):
         super(HierarchicalLogisticRegression, self).__init__()
         self.num_cats = None
 
+        self.mu_alpha_mu    = mu_alpha_mu
+        self.mu_alpha_sd    = mu_alpha_sd
+        self.sigma_alpha_sd = sigma_alpha_sd
+        self.mu_beta_mu     = mu_beta_mu
+        self.mu_beta_sd     = mu_beta_sd
+        self.sigma_beta_sd  = sigma_beta_sd
+
     def create_model(self):
         """
         Creates and returns the PyMC3 model.
@@ -42,11 +55,11 @@ def create_model(self):
         model = pm.Model()
 
         with model:
-            mu_alpha = pm.Normal('mu_alpha', mu=0, sd=100)
-            sigma_alpha = pm.HalfNormal('sigma_alpha', sd=100)
+            mu_alpha = pm.Normal('mu_alpha', mu=self.mu_alpha_mu, sd=self.mu_alpha_sd)
+            sigma_alpha = pm.HalfNormal('sigma_alpha', sd=self.sigma_alpha_sd)
 
-            mu_beta = pm.Normal('mu_beta', mu=0, sd=100)
-            sigma_beta = pm.HalfNormal('sigma_beta', sd=100)
+            mu_beta = pm.Normal('mu_beta', mu=self.mu_beta_mu, sd=self.mu_beta_sd)
+            sigma_beta = pm.HalfNormal('sigma_beta', sd=self.sigma_beta_sd)
 
             alpha = pm.Normal('alpha', mu=mu_alpha, sd=sigma_alpha, shape=(self.num_cats,))
             betas = pm.Normal('beta', mu=mu_beta, sd=sigma_beta, shape=(self.num_cats, self.num_pred))

pymc3_models/models/LinearRegression.py

@@ -13,9 +13,20 @@ class LinearRegression(BayesianModel):
     Linear Regression built using PyMC3.
     """
 
-    def __init__(self):
+    def __init__(self,
+                 alpha_mu = 0,
+                 alpha_sd = 100,
+                 beta_mu  = 0,
+                 beta_sd  = 100,
+                 tau      = 1):
         super(LinearRegression, self).__init__()
 
+        self.alpha_mu = alpha_mu
+        self.alpha_sd = alpha_sd
+        self.beta_mu  = beta_mu
+        self.beta_sd  = beta_sd
+        self.tau      = tau
+
     def create_model(self):
         """
         Creates and returns the PyMC3 model.
@@ -38,10 +49,10 @@ def create_model(self):
         model = pm.Model()
 
         with model:
-            alpha = pm.Normal('alpha', mu=0, sd=100, shape=(1))
-            betas = pm.Normal('betas', mu=0, sd=100, shape=(1, self.num_pred))
+            alpha = pm.Normal('alpha', mu=self.alpha_mu, sd=self.alpha_sd, shape=(1))
+            betas = pm.Normal('betas', mu=self.beta_mu, sd=self.beta_sd, shape=(1, self.num_pred))
 
-            s = pm.HalfNormal('s', tau=1)
+            s = pm.HalfNormal('s', tau=self.tau)
 
             mean = alpha + T.sum(betas * model_input, 1)
 
@@ -93,17 +104,17 @@ def fit(self, X, y, inference_type='advi', minibatch_size=None, inference_args=N
 
         return self
 
-    def predict(self, X, return_std=False):
+
+    def sample(self, X, samples=2000):
         """
-        Predicts values of new data with a trained Linear Regression model
+        samples the conditional posterior estimates
 
         Parameters
         ----------
         X : numpy array, shape [n_samples, n_features]
 
-        return_std : Boolean flag of whether to return standard deviations with mean values. Defaults to False.
+        samples : number of draws to make for each point
         """
-
         if self.trace is None:
             raise PyMC3ModelsError('Run fit on the model before predict.')
 
@@ -116,6 +127,23 @@ def predict(self, X, return_std=False):
 
         ppc = pm.sample_ppc(self.trace, model=self.cached_model, samples=2000)
 
+        return ppc
+
+    def predict(self, X, return_std=False, samples=2000):
+        """
+        Predicts values of new data with a trained Linear Regression model
+
+        Parameters
+        ----------
+        X : numpy array, shape [n_samples, n_features]
+
+        return_std : Boolean flag of whether to return standard deviations with mean values. Defaults to False.
+
+        samples: numberof draws to make for each input
+        """
+
+        ppc = self.sample(X, samples)
+
         if return_std:
             return ppc['y'].mean(axis=0), ppc['y'].std(axis=0)
         else: