Merge pull request #58 from eth-cscs/release-0.5.6

Release 0.5.6

README.md

@@ -6,7 +6,7 @@ algorithms and other likelihood-free inference schemes. It presently includes:
 
 * RejectionABC
 * PMCABC (Population Monte Carlo ABC)
-* SMCABC (Sequential Monte Carlo ABC)
+* SMCABC (Sequential Monte Carlo ABC) 
 * RSMCABC (Replenishment SMC-ABC)
 * APMCABC (Adaptive Population Monte Carlo ABC)
 * SABC (Simulated Annealing ABC)
@@ -26,9 +26,9 @@ scientists by providing
 # Documentation
 For more information, check out the
 
-* [Documentation](http://abcpy.readthedocs.io/en/v0.5.5) 
-* [Examples](https://github.com/eth-cscs/abcpy/tree/v0.5.5/examples) directory and
-* [Reference](http://abcpy.readthedocs.io/en/v0.5.5/abcpy.html)
+* [Documentation](http://abcpy.readthedocs.io/en/v0.5.6) 
+* [Examples](https://github.com/eth-cscs/abcpy/tree/v0.5.6/examples) directory and
+* [Reference](http://abcpy.readthedocs.io/en/v0.5.6/abcpy.html)
 
 
 Further, we provide a
@@ -55,7 +55,7 @@ finally CSCS (Swiss National Super Computing Center) for their generous support.
 
 There is a paper in the proceedings of the 2017 PASC conference. In case you use
 ABCpy for your publication, we would appreciate a citation. You can use
-[this](https://github.com/eth-cscs/abcpy/blob/v0.5.5/doc/literature/DuttaS-ABCpy-PASC-2017.bib)
+[this](https://github.com/eth-cscs/abcpy/blob/v0.5.6/doc/literature/DuttaS-ABCpy-PASC-2017.bib)
 
 BibTex reference.
 

VERSION

@@ -1 +1 @@
-0.5.5
+0.5.6

abcpy/acceptedparametersmanager.py

@@ -92,7 +92,8 @@ def get_mapping(self, models, is_root=True, index=0):
         Returns
         -------
         list
-            The first entry corresponds to the mapping of the root model, as well as all its parents. The second entry corresponds to the next index in depth-first search.
+            The first entry corresponds to the mapping of the root model, as well as all its parents. The second entry
+            corresponds to the next index in depth-first search.
         """
 
         # Implement a dfs to discover all nodes of the model
@@ -104,9 +105,9 @@ def get_mapping(self, models, is_root=True, index=0):
 
                 # Only parameters that are neither root nor Hyperparameters are included in the mapping
                 if(not(is_root) and not(isinstance(model, ModelResultingFromOperation))):
-                    for i in range(model.get_output_dimension()):
-                        mapping.append((model, index))
-                        index+=1
+                    #for i in range(model.get_output_dimension()):
+                    mapping.append((model, index))
+                    index+=1
 
                 for parent in model.get_input_models():
                     parent_mapping, index = self.get_mapping([parent], is_root= False, index=index)
@@ -139,13 +140,15 @@ def get_accepted_parameters_bds_values(self, models):
 
         # The self.accepted_parameters_bds.value() list has dimensions d x n_steps, where d is the number of free parameters
         accepted_bds_values = [[] for i in range(len(self.accepted_parameters_bds.value()))]
+
         # Add all columns that correspond to desired parameters to the list that is returned
         for model in models:
             for prob_model, index in mapping:
                 if(model==prob_model):
                     for i in range(len(self.accepted_parameters_bds.value())):
                         accepted_bds_values[i].append(self.accepted_parameters_bds.value()[i][index])
-        accepted_bds_values = [np.array(x).reshape(-1, ) for x in accepted_bds_values]
+        #accepted_bds_values = [np.array(x).reshape(-1, ) for x in accepted_bds_values]
+
         return accepted_bds_values
 
     def _reset_flags(self, models=None):

abcpy/continuousmodels.py

@@ -87,7 +87,7 @@ def forward_simulate(self, input_values, k, rng=np.random.RandomState(), mpi_com
         samples = np.zeros(shape=(k, self.get_output_dimension()))
         for j in range(0, self.get_output_dimension()):
             samples[:, j] = rng.uniform(input_values[j], input_values[j+self.get_output_dimension()], k)
-        return [np.array(x) for x in samples]
+        return [np.array(x).reshape(-1,) for x in samples]
 
 
     def get_output_dimension(self):
@@ -189,7 +189,7 @@ def forward_simulate(self, input_values, k, rng=np.random.RandomState(), mpi_com
         mu = input_values[0]
         sigma = input_values[1]
         result = np.array(rng.normal(mu, sigma, k))
-        return [np.array([x]) for x in result]
+        return [np.array([x]).reshape(-1,) for x in result]
 
 
     def get_output_dimension(self):
@@ -270,7 +270,7 @@ def forward_simulate(self, input_values, k, rng=np.random.RandomState(), mpi_com
         mean = input_values[0]
         df = input_values[1]
         result = np.array((rng.standard_t(df,k)+mean))
-        return [np.array([x]) for x in result]
+        return [np.array([x]).reshape(-1,) for x in result]
 
 
     def _check_input(self, input_values):
@@ -346,20 +346,13 @@ def __init__(self, parameters, name='Multivariate Normal'):
             raise TypeError('Input for Multivariate Normal has to be of type list.')
         if len(parameters)<2:
             raise ValueError('Input for Multivariate Normal has to be of length 2.')
-        if not isinstance(parameters[0], list):
-            raise TypeError('Input for mean of MultivarateNormal has to be of type list.')
-        if not isinstance(parameters[1], list):
-            raise TypeError('Input for covariance of MultivarateNormal has to be of type list.')
-
-        ## This part confuses me as you say mean may not be list!!
 
         mean = parameters[0]
         if isinstance(mean, list):
             self._dimension = len(mean)
-            input_parameters = InputConnector.from_list(parameters)
         elif isinstance(mean, ProbabilisticModel):
             self._dimension = mean.get_output_dimension()
-            input_parameters = parameters
+        input_parameters = InputConnector.from_list(parameters)
 
         super(MultivariateNormal, self).__init__(input_parameters, name)
         self.visited = False

abcpy/discretemodels.py

@@ -369,9 +369,8 @@ def forward_simulate(self, input_values, k, rng=np.random.RandomState()):
         list: [np.ndarray]
             A list containing the sampled values as np-array.
         """
-
-        result = np.array(rng.randint(input_values[0], input_values[1], size=k, dtype=np.int64))
-        return [np.array([x]) for x in result]
+        result = np.array(rng.randint(input_values[0], input_values[1]+1, size=k, dtype=np.int64))
+        return [np.array([x]).reshape(-1,) for x in result]
 
     def get_output_dimension(self):
         return self._dimension
@@ -391,8 +390,11 @@ def pmf(self, input_values, x):
         float:
             The pmf evaluated at point x.
         """
-        upperbound, lowerbound = input_values[0], input_values[1]
-        pmf = 1. / (upperbound - lowerbound + 1)
+        lowerbound, upperbound = input_values[0], input_values[1]
+        if x >= lowerbound and x <= upperbound:
+            pmf = 1. / (upperbound - lowerbound + 1)
+        else:
+            pmf = 0
         self.calculated_pmf = pmf
         return pmf
 

abcpy/graphtools.py

@@ -134,6 +134,7 @@ def _recursion_pdf_of_prior(self, models, parameters, mapping=None, is_root=True
         # At the beginning of calculation, obtain the mapping
         if(is_root):
             mapping, garbage_index = self._get_mapping()
+
         # The pdf of each root model is first calculated seperately
         result = [1.]*len(models)
 
@@ -146,9 +147,9 @@ def _recursion_pdf_of_prior(self, models, parameters, mapping=None, is_root=True
                 for mapped_model, model_index in mapping:
                     if(mapped_model==model):
                         parameter_index = model_index
-                        for j in range(model.get_output_dimension()):
-                            relevant_parameters.append(parameters[parameter_index])
-                            parameter_index+=1
+                        #for j in range(model.get_output_dimension()):
+                        relevant_parameters.append(parameters[parameter_index])
+                        #parameter_index+=1
                         break
                 if(len(relevant_parameters)==1):
                     relevant_parameters = relevant_parameters[0]
@@ -210,7 +211,7 @@ def _get_mapping(self, models=None, index=0, is_not_root=False):
             # If this model corresponds to an unvisited free parameter, add it to the mapping
             if(is_not_root and not(model.visited) and not(isinstance(model, Hyperparameter)) and not(isinstance(model, ModelResultingFromOperation))):
                 mapping.append((model, index))
-                index+=model.get_output_dimension()
+                index+= 1 #model.get_output_dimension()
             # Add all parents to the mapping, if applicable
             for parent in model.get_input_models():
                 parent_mapping, index = self._get_mapping([parent], index=index, is_not_root=True)
@@ -240,6 +241,7 @@ def _get_names_and_parameters(self):
         return_value = []
 
         for model, index in mapping:
+
             return_value.append((model.name, self.accepted_parameters_manager.get_accepted_parameters_bds_values([model])))
 
         return return_value
@@ -318,10 +320,11 @@ def set_parameters(self, parameters, models=None, index=0, is_root=True):
         for model in models:
             # New parameters should only be set in case we are not at the root
             if not is_root and not isinstance(model, ModelResultingFromOperation):
-                new_output_values = np.array(parameters[index:index + model.get_output_dimension()])
+                #new_output_values = np.array(parameters[index:index + model.get_output_dimension()])
+                new_output_values = np.array(parameters[index]).reshape(-1,)
                 if not model.set_output_values(new_output_values):
                     return [False, index]
-                index += model.get_output_dimension()
+                index += 1 #model.get_output_dimension()
                 model.visited = True
 
             # New parameters for all parents are set using a depth-first search