def __init__(self):
super(SimulationLearner, self).__init__()
# specification
self.specification = SimulationLearnerSpecification()
self.knowledge = ASGCKnowledge()
# statistics per knowledge type
self.trainingOverall = {}
self.trainAccuracy = {}
self.trainCount = {}
self.testAccuracy = {}
self.testingOverall = {}
self.testCount = {}
self.numberPoints = {}
self.level = {}
# initialize data container
self.dataContainer = {}
# there is one learner per time step
self._learners = {}
self._verbose = True
def __init__(self, builder):
self._builder = builder
self.__specification = SimulationLearnerSpecification()
self._builder.getSimulationLearner()\
.setSpecification(self.__specification)
class SimulationLearnerDescriptor(object):
"""
Specification descriptor for ASGCSampler
"""
def __init__(self, builder):
self._builder = builder
self.__specification = SimulationLearnerSpecification()
self._builder.getSimulationLearner()\
.setSpecification(self.__specification)
def withStartingIterationNumber(self, iteration):
"""
Set the starting iteration number
@param iteration: integer starting iteration number
"""
self._builder.getSimulationLearner().setCurrentIterationNumber(iteration)
return self
def withRefinement(self):
"""
Define if spatially adaptive refinement should be done and how...
"""
return RefinementDescriptor(self._builder)
def withParameters(self, params):
"""
Set the parameter setting
@param params: ParameterSet
"""
self.__specification.setParameters(params)
return self
def withQoI(self, qoi):
"""
Define, which quantity of interest we study.
@param qoi: string quantity of interest
"""
self.__specification.setQoI(qoi)
return self
def withTypesOfKnowledge(self, knowledgeTypes):
"""
Define for which type of functions the hierarchical coefficients are
computed using the specified learner.
@param knowledgeTypes: list of KnowledgetTypes
"""
self.__specification.setKnowledgeTypes(knowledgeTypes)
return self
def withTimeStepsOfInterest(self, ts):
"""
Define the time steps in which we are interested in. The learner
just learns those, which are specified here. Moreover, it considers
just this time steps for refinement.
@param ts: list of floats, time steps
"""
# check for uniqueness of time steps
if len(ts) != len(set(ts)):
raise AttributeError('time steps of interest are not unique')
self.__specification.setTimeStepsOfInterest(ts)
return self
def withAdaptThreshold(self, value):
"""
Specifies refinement threshold
@param value: float for refinement threshold
"""
self.__specification.setAdaptThreshold(value)
return self
def withAdaptPoints(self, value):
"""
Specifies number of points, which have to be refined in refinement step
@param value: integer for number of points to refine
"""
self.__specification.setAdaptPoints(value)
return self
def withAdaptRate(self, value):
"""
Specifies rate from total number of points on grid, which should be
refined.
@param value: float for rate
"""
self.__specification.setAdaptRate(value)
return self