Merge pull request #27 from eth-cscs/RemotecontextRefactoring

Remotecontext refactoring

abcpy/approx_lhd.py

@@ -6,27 +6,28 @@
 
 class Approx_likelihood(metaclass = ABCMeta):
     """This abstract base class defines the approximate likelihood 
-    function. To approximate the likelihood function at a parameter value given observed dataset, 
-    we need to pass a dataset simulated from model set at the parameter value and the observed dataset. 
+    function. To approximate the likelihood function at a parameter value given observed data set,
+    we need to pass a data set simulated from model set at the parameter value and the observed data set.
     """
 
     @abstractmethod
     def __init__(self, statistics_calc):       
-        """    The constructor of a sub-class must accept a non-optional statistics
-    calculator, which is stored to self.statistics_calc.
+        """
+        The constructor of a sub-class must accept a non-optional statistics
+        calculator, which is stored to self.statistics_calc.
 
-    Parameters
-    ----------
-    statistics_calc : abcpy.stasistics.Statistics 
-        Statistics extractor object that conforms to the Statistics class.
+        Parameters
+        ----------
+        statistics_calc : abcpy.stasistics.Statistics
+            Statistics extractor object that conforms to the Statistics class.
         """
 
         raise NotImplemented
 
     @abstractmethod
     def likelihood(y_obs, y_sim):
         """To be overwritten by any sub-class: should compute the approximate likelihood 
-        value given the observed dataset y_obs and dataset y_sim simulated from
+        value given the observed data set y_obs and the data set y_sim simulated from
         model set at the parameter value.
 
         Parameters
@@ -46,7 +47,7 @@ def likelihood(y_obs, y_sim):
 
 
 class SynLiklihood(Approx_likelihood):
-    """This class implements the aproximate likelihood function which computes the pproximate
+    """This class implements the approximate likelihood function which computes the approximate
     likelihood using the synthetic likelihood approach described in Wood [1].
     For synthetic likelihood approximation, we compute the robust precision matrix using Ledoit and Wolf's [2]
     method.
@@ -92,8 +93,8 @@ def likelihood(self, y_obs, y_sim):
 
 
 class PenLogReg(Approx_likelihood):
-    """This class implements the aproximate likelihood function which computes the pproximate
-    likelihood upto a constant using penalized logistic regression described in 
+    """This class implements the approximate likelihood function which computes the approximate
+    likelihood up to a constant using penalized logistic regression described in
     Dutta et. al. [1]. It takes one additional function handler defining the 
     true model and two additional parameters n_folds and n_simulate correspondingly defining number
     of folds used to estimate prediction error using cross-validation and the number 
@@ -106,10 +107,8 @@ class PenLogReg(Approx_likelihood):
     
     [2] Friedman, J., Hastie, T., and Tibshirani, R. (2010). Regularization 
     paths for generalized linear models via coordinate descent. Journal of Statistical 
-    Software, 33(1), 1–22.      
-    """
-    def __init__(self, statistics_calc, model, n_simulate, n_folds=10, max_iter = 100000, seed = None):
-        """
+    Software, 33(1), 1–22.
+
     Parameters
     ----------
     statistics_calc : abcpy.stasistics.Statistics
@@ -124,17 +123,18 @@ def __init__(self, statistics_calc, model, n_simulate, n_folds=10, max_iter = 10
         Maximum passes over the data. The default is 100000.
     seed: int, optional
         Seed for the random number generator. The used glmnet solver is not
-        deterministic, this seed is used for determining the cv folds. The default value is 
+        deterministic, this seed is used for determining the cv folds. The default value is
         None.
-        """
-
+    """
+    def __init__(self, statistics_calc, model, n_simulate, n_folds=10, max_iter = 100000, seed = None):
+
         self.model = model
         self.statistics_calc = statistics_calc
         self.n_folds = n_folds
         self.n_simulate = n_simulate
         self.seed = seed
         self.max_iter = max_iter
-        # Simulate reference data and extract summary statistics from the reffernce data      
+        # Simulate reference data and extract summary statistics from the reference data
         self.ref_data_stat = self._simulate_ref_data()
 
 
@@ -164,7 +164,7 @@ def likelihood(self, y_obs, y_sim):
 
     def _simulate_ref_data(self):
         """
-        Simulate the reference dataset. This code is run at the initializtion of 
+        Simulate the reference data set. This code is run at the initialization of
         Penlogreg
         """
 

abcpy/backends/base.py

@@ -30,7 +30,7 @@ def parallelize(self, list):
             A reference object that represents the parallelized list
         """
 
-        raise NotImplemented
+        raise NotImplementedError
 
 
     @abstractmethod
@@ -49,7 +49,7 @@ def broadcast(self, object):
             A reference to the broadcasted object
         """
 
-        raise NotImplemented
+        raise NotImplementedError
 
 
     @abstractmethod
@@ -72,7 +72,7 @@ def map(self, func, pds):
             a new parallel data set that contains the result of the map
         """
 
-        raise NotImplemented
+        raise NotImplementedError
 
 
     @abstractmethod
@@ -91,7 +91,7 @@ def collect(self, pds):
             all elements of pds as a list
         """
 
-        raise NotImplemented
+        raise NotImplementedError
 
 
 class PDS:
@@ -101,7 +101,7 @@ class PDS:
 
     @abstractmethod
     def __init__(self):
-        raise NotImplemented
+        raise NotImplementedError
 
 
 class BDS:
@@ -111,15 +111,15 @@ class BDS:
 
     @abstractmethod
     def __init__(self):
-        raise NotImplemented
+        raise NotImplementedError
 
 
     @abstractmethod
     def value(self):
         """
         This method should return the actual object that the broadcast data set represents. 
         """
-        raise NotImplemented
+        raise NotImplementedError
 
 
 class BackendDummy(Backend):

abcpy/backends/mpi.py

@@ -1,5 +1,6 @@
 import numpy as np
 import cloudpickle
+import pickle
 
 from mpi4py import MPI
 from abcpy.backends import Backend, PDS, BDS
@@ -54,7 +55,7 @@ def __command_slaves(self, command, data):
         elif command == self.OP_MAP:
             #In map we receive data as (pds_id,pds_id_new,func)
             #Use cloudpickle to dump the function into a string.
-            function_packed = self.__sanitize_and_pack_func(data[2])
+            function_packed = cloudpickle.dumps(data[2], pickle.HIGHEST_PROTOCOL)
             data_packet = (command, data[0], data[1], function_packed)
 
         elif command == self.OP_BROADCAST:
@@ -74,35 +75,6 @@ def __command_slaves(self, command, data):
         _ = self.comm.bcast(data_packet, root=0)
 
 
-    def __sanitize_and_pack_func(self, func):
-        """
-        Prevents the function from packing the backend by temporarily
-        setting it to another variable and then uses cloudpickle
-        to pack it into a string to be sent.
-
-        Parameters
-        ----------
-        func: Python Function
-            The function we are supposed to pack while sending it along to the slaves
-            during the map function
-
-        Returns
-        -------
-        Returns a string of the function packed by cloudpickle
-
-        """
-
-        #Set the backend to None to prevent it from being packed
-        globals()['backend'] = {}
-
-        function_packed = cloudpickle.dumps(func)
-
-        #Reset the backend to self after it's been packed
-        globals()['backend'] = self
-
-        return function_packed
-
-
     def __generate_new_pds_id(self):
         """
         This method generates a new pds_id to associate a PDS with it's remote counterpart

abcpy/distances.py

@@ -21,7 +21,7 @@ def __init__(self, statistics_calc):
             Statistics extractor object that conforms to the Statistics class.
         """
 
-        raise NotImplemented
+        raise NotImplementedError
 
 
     @abstractmethod
@@ -58,11 +58,11 @@ def distance(d1, d2):
             The distance between the two input data sets.
         """
 
-        raise NotImplemented
+        raise NotImplementedError
 
 
     @abstractmethod
-    def dist_max():
+    def dist_max(self):
         """To be overwritten by sub-class: should return maximum possible value of the
         desired distance function.
  
@@ -76,7 +76,7 @@ def dist_max():
             The maximal possible value of the desired distance function.
         """
 
-        raise NotImplemented
+        raise NotImplementedError
 
 
     def _calculate_summary_stat(self,d1,d2):

abcpy/distributions.py

@@ -19,12 +19,12 @@ def set_parameters(self, params):
 
         Parameters
         ----------
-        theta: list
+        params: list
             Contains all the distributions parameters.
             
         """
 
-        raise NotImplemented
+        raise NotImplementedError
 
 
 
@@ -39,7 +39,7 @@ def reseed(self, seed):
             
         """
 
-        raise NotImplemented
+        raise NotImplementedError
 
 
 
@@ -58,7 +58,7 @@ def sample(self, k):
             
         """
 
-        raise NotImplemented
+        raise NotImplementedError
 
 
     @abstractmethod
@@ -77,7 +77,7 @@ def pdf(self, x):
             
         """
 
-        raise NotImplemented
+        raise NotImplementedError
 
 
 class MultiNormal(Distribution):
@@ -130,7 +130,7 @@ class Uniform(Distribution):
     """
     This class implements a p-dimensional uniform Prior distribution in a closed interval.
     """    
-    def __init__(self, lb : np.ndarray, ub : np.ndarray, seed=None):
+    def __init__(self, lb, ub, seed=None):
         """
         Defines the upper and lower bounds of a p-dimensional uniform Prior distribution in a closed interval.