Merge pull request #379 from smattis/v3-steve

5/19 meeting

bet/calculateP/__init__.py

@@ -6,7 +6,7 @@
 
 * :mod:`~bet.calculateP.calculateP` provides methods for approximating probability densities in the measure-theoretic framework.
 * :mod:`~bet.calculateP.simpleFunP` provides methods for creating simple function approximations of probability densities for the measure-theoretic framework.
-* :mod:`~bet.calculateP.dataConsistent` provides methods for data-consistent stochastic inversion.
+* :mod:`~bet.calculateP.calculateR` provides methods for data-consistent stochastic inversion.
 * :mod:`~bet.calculateP.calculateError` provides methods for approximating numerical and sampling errors.
 """
-__all__ = ['calculateP', 'simpleFunP', 'calculateError', 'dataConsistent']
+__all__ = ['calculateP', 'simpleFunP', 'calculateError', 'calculateR']

bet/calculateP/dataConsistent.py → bet/calculateP/calculateR.py

@@ -1,14 +1,14 @@
 # Copyright (C) 2014-2020 The BET Development Team
 
 r"""
-This module contains functions for data-consistent stochastic inversion.
+This module contains functions for data-consistent stochastic inversion based on ratios of densities.
 
-* :meth:`~bet.calculateP.dataConsistent.generate_output_kdes` generates KDEs on output sets.
-* :meth:`~bet.calculateP.dataConsistent.invert_to_kde` solves SIP for weighted KDEs.
-* :meth:`~bet.calculateP.dataConsistent.invert_to_gmm` solves SIP for a Gaussian Mixture Model.
-* :meth:`~bet.calculateP.dataConsistent.invert_to_multivariate_gaussian` solves SIP for a multivariate Gaussian.
-* :meth:`~bet.calculateP.dataConsistent.invert_to_random_variable` solves SIP for random variables.
-* :meth:`~bet.calculateP.dataConsistent.invert_rejection_sampling` solves SIP with rejection sampling.
+* :meth:`~bet.calculateP.calculateR.generate_output_kdes` generates KDEs on output sets.
+* :meth:`~bet.calculateP.calculateR.invert_to_kde` solves SIP for weighted KDEs.
+* :meth:`~bet.calculateP.calculateR.invert_to_gmm` solves SIP for a Gaussian Mixture Model.
+* :meth:`~bet.calculateP.calculateR.invert_to_multivariate_gaussian` solves SIP for a multivariate Gaussian.
+* :meth:`~bet.calculateP.calculateR.invert_to_random_variable` solves SIP for random variables.
+* :meth:`~bet.calculateP.calculateR.invert_rejection_sampling` solves SIP with rejection sampling.
 
 """
 import bet.sample
@@ -33,13 +33,13 @@ def generate_output_kdes(discretization, bw_method=None):
     discretization.local_to_global()
 
     predict_set = discretization.get_output_sample_set()
-    obs_set = discretization.get_output_probability_set()
+    obs_set = discretization.get_output_observed_set()
     if predict_set.get_region() is None or obs_set.get_region() is None:
         predict_set.set_region(np.array([0] * predict_set.check_num()))
         obs_set.set_region(np.array([0] * obs_set.check_num()))
 
     if predict_set.get_cluster_maps() is None:
-        num_clusters = int(max(np.max(predict_set.get_region()), np.max(obs_set.get_region())) + 1)
+        num_clusters = int(np.max(predict_set.get_region()) + 1)
     else:
         num_clusters = len(predict_set.get_cluster_maps())
 
@@ -80,7 +80,7 @@ def generate_output_kdes(discretization, bw_method=None):
                 obs_kdes.append(gaussian_kde(obs_set.get_values()[obs_pointer].T, bw_method=bw_method))
             else:
                 obs_kdes.append(None)
-    return predict_set, predict_kdes, obs_set, obs_kdes, num_clusters
+    return predict_kdes, obs_kdes, num_clusters
 
 
 def invert(discretization, bw_method = None):
@@ -96,7 +96,8 @@ def invert(discretization, bw_method = None):
     :return: marginal probabilities and cluster weights
     :rtype: list, `np.ndarray`
     """
-    predict_set, predict_kdes, obs_set, obs_kdes, num_clusters = generate_output_kdes(discretization, bw_method)
+    predict_kdes, obs_kdes, num_clusters = generate_output_kdes(discretization, bw_method)
+    predict_set = discretization.get_output_sample_set()
 
     rs = []
     r = []
@@ -136,10 +137,12 @@ def invert_to_kde(discretization, bw_method = None):
     """
     from scipy.stats import gaussian_kde
 
-    predict_set, predict_kdes, obs_set, obs_kdes, num_clusters = generate_output_kdes(discretization, bw_method)
+    predict_kdes, obs_kdes, num_clusters = generate_output_kdes(discretization, bw_method)
 
     rs, r, lam_ptr = invert(discretization, bw_method)
 
+    obs_set = discretization.get_output_observed_set()
+
     # Compute marginal probabilities for each parameter and initial condition.
     param_marginals = []
     cluster_weights = []
@@ -176,8 +179,7 @@ def invert_rejection_sampling(discretization, bw_method=None):
     :return: sample set containing samples
     :rtype: :class:`bet.sample.sample_set`
     """
-    predict_set, predict_kdes, obs_set, obs_kdes, num_clusters = generate_output_kdes(discretization,
-                                                                                      bw_method=bw_method)
+    predict_kdes, obs_kdes, num_clusters = generate_output_kdes(discretization, bw_method)
 
     rs, r, lam_ptr = invert(discretization, bw_method)
 
@@ -233,10 +235,12 @@ def weighted_mean_and_cov(x, weights):
         cov1 = cov1 / sum_weights
         return mean1, cov1
 
-    predict_set, predict_kdes, obs_set, obs_kdes, num_clusters = generate_output_kdes(discretization, bw_method)
+    predict_kdes, obs_kdes, num_clusters = generate_output_kdes(discretization, bw_method)
 
     rs, r, lam_ptr = invert(discretization, bw_method)
 
+    obs_set = discretization.get_output_observed_set()
+
     # Compute multivariate normal for each cluster
     means = []
     covariances = []
@@ -289,10 +293,12 @@ def weighted_mean_and_cov(x, weights):
         cov1 = cov1 / sum_weights
         return mean1, cov1
 
-    predict_set, predict_kdes, obs_set, obs_kdes, num_clusters = generate_output_kdes(discretization, bw_method)
+    predict_kdes, obs_kdes, num_clusters = generate_output_kdes(discretization, bw_method)
 
     rs, r, lam_ptr = invert(discretization, bw_method)
 
+    obs_set = discretization.get_output_observed_set()
+
     # Compute multivariate normal
     cluster_weights = []
     num_obs = obs_set.check_num()
@@ -358,10 +364,12 @@ def invert_to_random_variable(discretization, rv, num_reweighted=10000, bw_metho
     else:
         raise bet.sample.wrong_input("rv must be a string, list, or tuple.")
 
-    predict_set, predict_kdes, obs_set, obs_kdes, num_clusters = generate_output_kdes(discretization, bw_method)
+    predict_kdes, obs_kdes, num_clusters = generate_output_kdes(discretization, bw_method)
 
     rs, r, lam_ptr = invert(discretization, bw_method)
 
+    obs_set = discretization.get_output_observed_set()
+
     # Compute multivariate normal
     cluster_weights = []
     num_obs = obs_set.check_num()

bet/postProcess/compareP.py

@@ -132,9 +132,9 @@ def evaluate_pdfs(self):
         sup2 = np.equal(self.pdfs2, 0.0)
         self.pdfs_zero = np.sum(np.logical_and(sup1, sup2))
 
-    def distance(self, functional='tv', normalize=True, **kwargs):
+    def distance(self, functional='tv', normalize=False, **kwargs):
         """
-        Compute the discrete statistical distance between the probability measures
+        Compute the statistical distance between the probability measures
         evaluated at the comparison points.
 
         :param functional: functional defining type of statistical distance
@@ -183,10 +183,10 @@ def distance(self, functional='tv', normalize=True, **kwargs):
             dist = functional(self.pdfs1, self.pdfs2, **kwargs)
         return dist
 
-    def distance_marginal(self, i, interval=None, num_points=1000, compare_factor=0.0, normalize=True,
+    def distance_marginal(self, i, interval=None, num_points=1000, compare_factor=0.0, normalize=False,
                           functional='tv', **kwargs):
         """
-        Compute the discrete statistical distance between the marginals of the probability measures
+        Compute the statistical distance between the marginals of the probability measures
         evaluated at equally spaced points on an interval. If the interval is not defined,
         one is computed by the maximum and minimum values. This domain is extended by the proportion
         set by `compare_factor`.

bet/sample.py

@@ -2238,12 +2238,13 @@ class discretization(object):
     #: :class:`sample.sample_set_base`
     sample_set_names = ['_input_sample_set', '_output_sample_set',
                         '_emulated_input_sample_set', '_emulated_output_sample_set',
-                        '_output_probability_set']
+                        '_output_probability_set', '_output_observed_set']
 
     def __init__(self, input_sample_set, output_sample_set,
                  output_probability_set=None,
                  emulated_input_sample_set=None,
-                 emulated_output_sample_set=None):
+                 emulated_output_sample_set=None,
+                 output_observed_set=None):
         """
         Initialize the discretization.
         
@@ -2257,6 +2258,8 @@ def __init__(self, input_sample_set, output_sample_set,
         :type emulated_input_sample_set: :class:`bet.sample.sample_set_base`
         :param emulated_output_sample_set: Emulated output set
         :type emulated_output_sample_set: :class:`bet.sample.sample_set_base`
+        :param output_observed_set: Observed output set
+        :type output_observed_set: :class:`bet.sample.sample_set_base`
 
         """
         #: Input sample set :class:`~bet.sample.sample_set_base`
@@ -2271,6 +2274,8 @@ def __init__(self, input_sample_set, output_sample_set,
         self._output_probability_set = output_probability_set
         #: Pointer from ``self._output_sample_set`` to
         #: ``self._output_probability_set``
+        #: Observed output sample set :class:`~bet.sample.sample_set_base`
+        self._output_observed_set = output_observed_set
         self._io_ptr = None
         #: Pointer from ``self._emulated_input_sample_set`` to
         #: ``self._input_sample_set``
@@ -2533,6 +2538,31 @@ def set_output_sample_set(self, output_sample_set):
         else:
             raise AttributeError("Wrong Type: Should be sample_set_base type")
 
+    def get_output_observed_set(self):
+        """
+
+        Returns a reference to the output observed sample set for this discretization.
+
+        :rtype: :class:`~bet.sample.sample_set_base`
+        :returns: output sample set
+
+        """
+        return self._output_observed_set
+
+    def set_output_observed_set(self, output_sample_set):
+        """
+
+        Sets the output observed sample set for this discretization.
+
+        :param output_sample_set: output observed sample set.
+        :type output_sample_set: :class:`~bet.sample.sample_set_base`
+
+        """
+        if isinstance(output_sample_set, sample_set_base):
+            self._output_observed_set = output_sample_set
+        else:
+            raise AttributeError("Wrong Type: Should be sample_set_base type")
+
     def get_output_probability_set(self):
         """
 

bet/sampling/useLUQ.py

@@ -143,7 +143,7 @@ def make_disc(self):
         # Prediction discretization
         disc1 = sample.discretization(input_sample_set=self.predict_set,
                                       output_sample_set=predict_output,
-                                      output_probability_set=obs_output)
+                                      output_observed_set=obs_output)
 
         # Observation discretization
         disc2 = sample.discretization(input_sample_set=self.obs_set,

doc/bet.calculateP.rst

@@ -9,7 +9,6 @@ bet.calculateP.calculateError module
 
 .. automodule:: bet.calculateP.calculateError
     :members:
-    :special-members:
     :undoc-members:
     :show-inheritance:
 
@@ -18,16 +17,14 @@ bet.calculateP.calculateP module
 
 .. automodule:: bet.calculateP.calculateP
     :members:
-    :special-members:
     :undoc-members:
     :show-inheritance:
 
-bet.calculateP.dataConsistent module
+bet.calculateP.calculateR module
 ------------------------------------
 
-.. automodule:: bet.calculateP.dataConsistent
+.. automodule:: bet.calculateP.calculateR
     :members:
-    :special-members:
     :undoc-members:
     :show-inheritance:
 
@@ -36,7 +33,6 @@ bet.calculateP.simpleFunP module
 
 .. automodule:: bet.calculateP.simpleFunP
     :members:
-    :special-members:
     :undoc-members:
     :show-inheritance:
 

doc/bet.postProcess.rst

@@ -9,7 +9,6 @@ bet.postProcess.compareP module
 
 .. automodule:: bet.postProcess.compareP
     :members:
-    :special-members:
     :undoc-members:
     :show-inheritance:
 
@@ -18,7 +17,6 @@ bet.postProcess.plotDomains module
 
 .. automodule:: bet.postProcess.plotDomains
     :members:
-    :special-members:
     :undoc-members:
     :show-inheritance:
 
@@ -27,7 +25,6 @@ bet.postProcess.plotP module
 
 .. automodule:: bet.postProcess.plotP
     :members:
-    :special-members:
     :undoc-members:
     :show-inheritance:
 
@@ -36,7 +33,6 @@ bet.postProcess.plotVoronoi module
 
 .. automodule:: bet.postProcess.plotVoronoi
     :members:
-    :special-members:
     :undoc-members:
     :show-inheritance:
 
@@ -45,7 +41,6 @@ bet.postProcess.postTools module
 
 .. automodule:: bet.postProcess.postTools
     :members:
-    :special-members:
     :undoc-members:
     :show-inheritance:
 

doc/bet.rst

@@ -19,7 +19,6 @@ bet.Comm module
 
 .. automodule:: bet.Comm
     :members:
-    :special-members:
     :undoc-members:
     :show-inheritance:
 
@@ -28,7 +27,6 @@ bet.sample module
 
 .. automodule:: bet.sample
     :members:
-    :special-members:
     :undoc-members:
     :show-inheritance:
 
@@ -37,7 +35,6 @@ bet.surrogates module
 
 .. automodule:: bet.surrogates
     :members:
-    :special-members:
     :undoc-members:
     :show-inheritance:
 
@@ -46,7 +43,6 @@ bet.util module
 
 .. automodule:: bet.util
     :members:
-    :special-members:
     :undoc-members:
     :show-inheritance:
 

doc/bet.sampling.rst

@@ -9,7 +9,6 @@ bet.sampling.LpGeneralizedSamples module
 
 .. automodule:: bet.sampling.LpGeneralizedSamples
     :members:
-    :special-members:
     :undoc-members:
     :show-inheritance:
 
@@ -18,7 +17,6 @@ bet.sampling.basicSampling module
 
 .. automodule:: bet.sampling.basicSampling
     :members:
-    :special-members:
     :undoc-members:
     :show-inheritance:
 
@@ -27,7 +25,6 @@ bet.sampling.useLUQ module
 
 .. automodule:: bet.sampling.useLUQ
     :members:
-    :special-members:
     :undoc-members:
     :show-inheritance:
 

doc/bet.sensitivity.rst

@@ -9,7 +9,6 @@ bet.sensitivity.chooseQoIs module
 
 .. automodule:: bet.sensitivity.chooseQoIs
     :members:
-    :special-members:
     :undoc-members:
     :show-inheritance:
 
@@ -18,7 +17,6 @@ bet.sensitivity.gradients module
 
 .. automodule:: bet.sensitivity.gradients
     :members:
-    :special-members:
     :undoc-members:
     :show-inheritance:
 

doc/overview.rst

@@ -152,7 +152,7 @@ The package layout is as follows::
       calculateP 
       calculateError
       simpleFunP
-      dataConsistent
+      calculateR
     sampling/
       basicSampling
       useLUQ