# -*- coding: utf-8 -*-
#
-# Copyright (C) 2008-2020 EDF R&D
+# Copyright (C) 2008-2021 EDF R&D
#
# This library is free software; you can redistribute it and/or
# modify it under the terms of the GNU Lesser General Public
import sys
import logging
import copy
+import time
import numpy
from functools import partial
from daCore import Persistence, PlatformInfo, Interfaces
def clearCache(self):
"Vide le cache"
- self.__listOPCV = [] # Previous Calculated Points, Results, Point Norms, Operator
+ self.__listOPCV = []
self.__seenNames = []
- # logging.debug("CM Tolerance de determination des doublons : %.2e", self.__tolerBP)
- def wasCalculatedIn(self, xValue, oName="" ): #, info="" ):
+ def wasCalculatedIn(self, xValue, oName="" ):
"Vérifie l'existence d'un calcul correspondant à la valeur"
__alc = False
__HxV = None
if self.__enabled:
for i in range(min(len(self.__listOPCV),self.__lenghtOR)-1,-1,-1):
- if not hasattr(xValue, 'size') or (str(oName) != self.__listOPCV[i][3]) or (xValue.size != self.__listOPCV[i][0].size):
- # logging.debug("CM Différence de la taille %s de X et de celle %s du point %i déjà calculé", xValue.shape,i,self.__listOPCP[i].shape)
+ if not hasattr(xValue, 'size'):
pass
- elif numpy.linalg.norm(numpy.ravel(xValue) - self.__listOPCV[i][0]) < self.__tolerBP * self.__listOPCV[i][2]:
+ elif (str(oName) != self.__listOPCV[i][3]):
+ pass
+ elif (xValue.size != self.__listOPCV[i][0].size):
+ pass
+ elif (numpy.ravel(xValue)[0] - self.__listOPCV[i][0][0]) > (self.__tolerBP * self.__listOPCV[i][2] / self.__listOPCV[i][0].size):
+ pass
+ elif numpy.linalg.norm(numpy.ravel(xValue) - self.__listOPCV[i][0]) < (self.__tolerBP * self.__listOPCV[i][2]):
__alc = True
__HxV = self.__listOPCV[i][1]
- # logging.debug("CM Cas%s déja calculé, portant le numéro %i", info, i)
break
return __alc, __HxV
def storeValueInX(self, xValue, HxValue, oName="" ):
"Stocke pour un opérateur o un calcul Hx correspondant à la valeur x"
if self.__lenghtOR < 0:
- self.__lenghtOR = 2 * xValue.size + 2
+ self.__lenghtOR = 2 * min(xValue.size, 50) + 2 # 2 * xValue.size + 2
self.__initlnOR = self.__lenghtOR
self.__seenNames.append(str(oName))
if str(oName) not in self.__seenNames: # Etend la liste si nouveau
- self.__lenghtOR += 2 * xValue.size + 2
+ self.__lenghtOR += 2 * min(xValue.size, 50) + 2 # 2 * xValue.size + 2
self.__initlnOR += self.__lenghtOR
self.__seenNames.append(str(oName))
while len(self.__listOPCV) > self.__lenghtOR:
- # logging.debug("CM Réduction de la liste des cas à %i éléments par suppression du premier", self.__lenghtOR)
self.__listOPCV.pop(0)
self.__listOPCV.append( (
- copy.copy(numpy.ravel(xValue)),
- copy.copy(HxValue),
- numpy.linalg.norm(xValue),
- str(oName),
+ copy.copy(numpy.ravel(xValue)), # 0 Previous point
+ copy.copy(HxValue), # 1 Previous value
+ numpy.linalg.norm(xValue), # 2 Norm
+ str(oName), # 3 Operator name
) )
def disable(self):
"Renvoie le type"
return self.__Type
- def appliedTo(self, xValue, HValue = None, argsAsSerie = False):
+ def appliedTo(self, xValue, HValue = None, argsAsSerie = False, returnSerieAsArrayMatrix = False):
"""
Permet de restituer le résultat de l'application de l'opérateur à une
série d'arguments xValue. Cette méthode se contente d'appliquer, chaque
#
if _HValue is not None:
assert len(_xValue) == len(_HValue), "Incompatible number of elements in xValue and HValue"
- HxValue = []
+ _HxValue = []
for i in range(len(_HValue)):
- HxValue.append( numpy.asmatrix( numpy.ravel( _HValue[i] ) ).T )
+ _HxValue.append( numpy.asmatrix( numpy.ravel( _HValue[i] ) ).T )
if self.__AvoidRC:
- Operator.CM.storeValueInX(_xValue[i],HxValue[-1],self.__name)
+ Operator.CM.storeValueInX(_xValue[i],_HxValue[-1],self.__name)
else:
- HxValue = []
+ _HxValue = []
_xserie = []
_hindex = []
for i, xv in enumerate(_xValue):
else:
if self.__Matrix is not None:
self.__addOneMatrixCall()
- _hv = self.__Matrix * xv
+ _xv = numpy.matrix(numpy.ravel(xv)).T
+ _hv = self.__Matrix * _xv
else:
self.__addOneMethodCall()
_xserie.append( xv )
_hindex.append( i )
_hv = None
- HxValue.append( _hv )
+ _HxValue.append( _hv )
#
if len(_xserie)>0 and self.__Matrix is None:
if self.__extraArgs is None:
for i in _hindex:
_xv = _xserie.pop(0)
_hv = _hserie.pop(0)
- HxValue[i] = _hv
+ _HxValue[i] = _hv
if self.__AvoidRC:
Operator.CM.storeValueInX(_xv,_hv,self.__name)
#
- if argsAsSerie: return HxValue
- else: return HxValue[-1]
+ if returnSerieAsArrayMatrix:
+ _HxValue = numpy.stack([numpy.ravel(_hv) for _hv in _HxValue], axis=1)
+ #
+ if argsAsSerie: return _HxValue
+ else: return _HxValue[-1]
- def appliedControledFormTo(self, paires, argsAsSerie = False):
+ def appliedControledFormTo(self, paires, argsAsSerie = False, returnSerieAsArrayMatrix = False):
"""
Permet de restituer le résultat de l'application de l'opérateur à des
paires (xValue, uValue). Cette méthode se contente d'appliquer, son
PlatformInfo.isIterable( _xuValue, True, " in Operator.appliedControledFormTo" )
#
if self.__Matrix is not None:
- HxValue = []
+ _HxValue = []
for paire in _xuValue:
_xValue, _uValue = paire
+ _xValue = numpy.matrix(numpy.ravel(_xValue)).T
self.__addOneMatrixCall()
- HxValue.append( self.__Matrix * _xValue )
+ _HxValue.append( self.__Matrix * _xValue )
else:
- HxValue = []
+ _xuArgs = []
for paire in _xuValue:
- _xuValue = []
_xValue, _uValue = paire
if _uValue is not None:
- _xuValue.append( paire )
+ _xuArgs.append( paire )
else:
- _xuValue.append( _xValue )
- self.__addOneMethodCall( len(_xuValue) )
+ _xuArgs.append( _xValue )
+ self.__addOneMethodCall( len(_xuArgs) )
if self.__extraArgs is None:
- HxValue = self.__Method( _xuValue ) # Calcul MF
+ _HxValue = self.__Method( _xuArgs ) # Calcul MF
else:
- HxValue = self.__Method( _xuValue, self.__extraArgs ) # Calcul MF
+ _HxValue = self.__Method( _xuArgs, self.__extraArgs ) # Calcul MF
#
- if argsAsSerie: return HxValue
- else: return HxValue[-1]
+ if returnSerieAsArrayMatrix:
+ _HxValue = numpy.stack([numpy.ravel(_hv) for _hv in _HxValue], axis=1)
+ #
+ if argsAsSerie: return _HxValue
+ else: return _HxValue[-1]
- def appliedInXTo(self, paires, argsAsSerie = False):
+ def appliedInXTo(self, paires, argsAsSerie = False, returnSerieAsArrayMatrix = False):
"""
Permet de restituer le résultat de l'application de l'opérateur à une
série d'arguments xValue, sachant que l'opérateur est valable en
PlatformInfo.isIterable( _nxValue, True, " in Operator.appliedInXTo" )
#
if self.__Matrix is not None:
- HxValue = []
+ _HxValue = []
for paire in _nxValue:
_xNominal, _xValue = paire
+ _xValue = numpy.matrix(numpy.ravel(_xValue)).T
self.__addOneMatrixCall()
- HxValue.append( self.__Matrix * _xValue )
+ _HxValue.append( self.__Matrix * _xValue )
else:
self.__addOneMethodCall( len(_nxValue) )
if self.__extraArgs is None:
- HxValue = self.__Method( _nxValue ) # Calcul MF
+ _HxValue = self.__Method( _nxValue ) # Calcul MF
else:
- HxValue = self.__Method( _nxValue, self.__extraArgs ) # Calcul MF
+ _HxValue = self.__Method( _nxValue, self.__extraArgs ) # Calcul MF
+ #
+ if returnSerieAsArrayMatrix:
+ _HxValue = numpy.stack([numpy.ravel(_hv) for _hv in _HxValue], axis=1)
#
- if argsAsSerie: return HxValue
- else: return HxValue[-1]
+ if argsAsSerie: return _HxValue
+ else: return _HxValue[-1]
def asMatrix(self, ValueForMethodForm = "UnknownVoidValue", argsAsSerie = False):
"""
centeredDF = __Function["CenteredFiniteDifference"],
increment = __Function["DifferentialIncrement"],
dX = __Function["withdX"],
+ extraArguments = self.__extraArgs,
avoidingRedundancy = __Function["withAvoidingRedundancy"],
toleranceInRedundancy = __Function["withToleranceInRedundancy"],
lenghtOfRedundancy = __Function["withLenghtOfRedundancy"],
- OMB : Observation moins Background : Y - Xb
- ForecastState : état prédit courant lors d'itérations
- Residu : dans le cas des algorithmes de vérification
+ - SampledStateForQuantiles : échantillons d'états pour l'estimation des quantiles
- SigmaBck2 : indicateur de correction optimale des erreurs d'ébauche
- SigmaObs2 : indicateur de correction optimale des erreurs d'observation
- SimulatedObservationAtBackground : l'état observé H(Xb) à l'ébauche
self.StoredVariables["CostFunctionJbAtCurrentOptimum"] = Persistence.OneScalar(name = "CostFunctionJbAtCurrentOptimum")
self.StoredVariables["CostFunctionJo"] = Persistence.OneScalar(name = "CostFunctionJo")
self.StoredVariables["CostFunctionJoAtCurrentOptimum"] = Persistence.OneScalar(name = "CostFunctionJoAtCurrentOptimum")
- self.StoredVariables["CurrentIterationNumber"] = Persistence.OneIndex(name = "CurrentIterationNumber")
+ self.StoredVariables["CurrentEnsembleState"] = Persistence.OneMatrix(name = "CurrentEnsembleState")
+ self.StoredVariables["CurrentIterationNumber"] = Persistence.OneIndex(name = "CurrentIterationNumber")
self.StoredVariables["CurrentOptimum"] = Persistence.OneVector(name = "CurrentOptimum")
self.StoredVariables["CurrentState"] = Persistence.OneVector(name = "CurrentState")
self.StoredVariables["ForecastState"] = Persistence.OneVector(name = "ForecastState")
self.StoredVariables["OMA"] = Persistence.OneVector(name = "OMA")
self.StoredVariables["OMB"] = Persistence.OneVector(name = "OMB")
self.StoredVariables["Residu"] = Persistence.OneScalar(name = "Residu")
+ self.StoredVariables["SampledStateForQuantiles"] = Persistence.OneMatrix(name = "SampledStateForQuantiles")
self.StoredVariables["SigmaBck2"] = Persistence.OneScalar(name = "SigmaBck2")
self.StoredVariables["SigmaObs2"] = Persistence.OneScalar(name = "SigmaObs2")
self.StoredVariables["SimulatedObservationAtBackground"] = Persistence.OneVector(name = "SimulatedObservationAtBackground")
"Pré-calcul"
logging.debug("%s Lancement", self._name)
logging.debug("%s Taille mémoire utilisée de %.0f Mio"%(self._name, self._m.getUsedMemory("Mio")))
+ self._getTimeState(reset=True)
#
# Mise a jour des paramètres internes avec le contenu de Parameters, en
# reprenant les valeurs par défauts pour toutes celles non définies
__test_ovalue( EM, "EM", "Evolution", "M" )
__test_ovalue( CM, "CM", "Control Model", "C" )
#
+ # Corrections et compléments des bornes
if ("Bounds" in self._parameters) and isinstance(self._parameters["Bounds"], (list, tuple)) and (len(self._parameters["Bounds"]) > 0):
logging.debug("%s Prise en compte des bornes effectuee"%(self._name,))
else:
self._parameters["Bounds"] = None
#
+ # Corrections et compléments de l'initialisation en X
+ if "InitializationPoint" in self._parameters:
+ if Xb is not None:
+ if self._parameters["InitializationPoint"] is not None and hasattr(self._parameters["InitializationPoint"],'size'):
+ if self._parameters["InitializationPoint"].size != numpy.ravel(Xb).size:
+ raise ValueError("Incompatible size %i of forced initial point that have to replace the background of size %i" \
+ %(self._parameters["InitializationPoint"].size,numpy.ravel(Xb).size))
+ # Obtenu par typecast : numpy.ravel(self._parameters["InitializationPoint"])
+ else:
+ self._parameters["InitializationPoint"] = numpy.ravel(Xb)
+ else:
+ if self._parameters["InitializationPoint"] is None:
+ raise ValueError("Forced initial point can not be set without any given Background or required value")
+ #
+ # Correction pour pallier a un bug de TNC sur le retour du Minimum
+ if "Minimizer" in self._parameters and self._parameters["Minimizer"] == "TNC":
+ self.setParameterValue("StoreInternalVariables",True)
+ #
+ # Verbosité et logging
if logging.getLogger().level < logging.WARNING:
self._parameters["optiprint"], self._parameters["optdisp"] = 1, 1
if PlatformInfo.has_scipy:
logging.debug("%s Nombre d'évaluation(s) de l'opérateur d'observation direct/tangent/adjoint.: %i/%i/%i", self._name, _oH["Direct"].nbcalls(0),_oH["Tangent"].nbcalls(0),_oH["Adjoint"].nbcalls(0))
logging.debug("%s Nombre d'appels au cache d'opérateur d'observation direct/tangent/adjoint..: %i/%i/%i", self._name, _oH["Direct"].nbcalls(3),_oH["Tangent"].nbcalls(3),_oH["Adjoint"].nbcalls(3))
logging.debug("%s Taille mémoire utilisée de %.0f Mio", self._name, self._m.getUsedMemory("Mio"))
+ logging.debug("%s Durées d'utilisation CPU de %.1fs et elapsed de %.1fs", self._name, self._getTimeState()[0], self._getTimeState()[1])
logging.debug("%s Terminé", self._name)
return 0
"""
raise NotImplementedError("Mathematical assimilation calculation has not been implemented!")
- def defineRequiredParameter(self, name = None, default = None, typecast = None, message = None, minval = None, maxval = None, listval = None):
+ def defineRequiredParameter(self, name = None, default = None, typecast = None, message = None, minval = None, maxval = None, listval = None, listadv = None):
"""
Permet de définir dans l'algorithme des paramètres requis et leurs
caractéristiques par défaut.
"minval" : minval,
"maxval" : maxval,
"listval" : listval,
+ "listadv" : listadv,
"message" : message,
}
self.__canonical_parameter_name[name.lower()] = name
minval = self.__required_parameters[__k]["minval"]
maxval = self.__required_parameters[__k]["maxval"]
listval = self.__required_parameters[__k]["listval"]
+ listadv = self.__required_parameters[__k]["listadv"]
#
if value is None and default is None:
__val = None
raise ValueError("The parameter named '%s' of value '%s' can not be less than %s."%(__k, __val, minval))
if maxval is not None and (numpy.array(__val, float) > maxval).any():
raise ValueError("The parameter named '%s' of value '%s' can not be greater than %s."%(__k, __val, maxval))
- if listval is not None:
+ if listval is not None or listadv is not None:
if typecast is list or typecast is tuple or isinstance(__val,list) or isinstance(__val,tuple):
for v in __val:
- if v not in listval:
+ if listval is not None and v in listval: continue
+ elif listadv is not None and v in listadv: continue
+ else:
raise ValueError("The value '%s' is not allowed for the parameter named '%s', it has to be in the list %s."%(v, __k, listval))
- elif __val not in listval:
+ elif not (listval is not None and __val in listval) and not (listadv is not None and __val in listadv):
raise ValueError("The value '%s' is not allowed for the parameter named '%s', it has to be in the list %s."%( __val, __k,listval))
#
return __val
pass
logging.debug("%s %s : %s", self._name, self.__required_parameters[k]["message"], self._parameters[k])
+ def _getTimeState(self, reset=False):
+ """
+ Initialise ou restitue le temps de calcul (cpu/elapsed) à la seconde
+ """
+ if reset:
+ self.__initial_cpu_time = time.process_time()
+ self.__initial_elapsed_time = time.perf_counter()
+ return 0., 0.
+ else:
+ self.__cpu_time = time.process_time() - self.__initial_cpu_time
+ self.__elapsed_time = time.perf_counter() - self.__initial_elapsed_time
+ return self.__cpu_time, self.__elapsed_time
+
+ def _StopOnTimeLimit(self, X=None, withReason=False):
+ "Stop criteria on time limit: True/False [+ Reason]"
+ c, e = self._getTimeState()
+ if "MaximumCpuTime" in self._parameters and c > self._parameters["MaximumCpuTime"]:
+ __SC, __SR = True, "Reached maximum CPU time (%.1fs > %.1fs)"%(c, self._parameters["MaximumCpuTime"])
+ elif "MaximumElapsedTime" in self._parameters and e > self._parameters["MaximumElapsedTime"]:
+ __SC, __SR = True, "Reached maximum elapsed time (%.1fs > %.1fs)"%(e, self._parameters["MaximumElapsedTime"])
+ else:
+ __SC, __SR = False, ""
+ if withReason:
+ return __SC, __SR
+ else:
+ return __SC
+
# ==============================================================================
class AlgorithmAndParameters(object):
"""
__Matrix, __Scalar, __Vector, __Object = asCovariance, asEyeByScalar, asEyeByVector, asCovObject
#
if __Scalar is not None:
- if numpy.matrix(__Scalar).size != 1:
- raise ValueError(' The diagonal multiplier given to define a sparse matrix is not a unique scalar value.\n Its actual measured size is %i. Please check your scalar input.'%numpy.matrix(__Scalar).size)
+ if isinstance(__Scalar, str):
+ __Scalar = __Scalar.replace(";"," ").replace(","," ").split()
+ if len(__Scalar) > 0: __Scalar = __Scalar[0]
+ if numpy.array(__Scalar).size != 1:
+ raise ValueError(' The diagonal multiplier given to define a sparse matrix is not a unique scalar value.\n Its actual measured size is %i. Please check your scalar input.'%numpy.array(__Scalar).size)
self.__is_scalar = True
self.__C = numpy.abs( float(__Scalar) )
self.shape = (0,0)
self.size = 0
elif __Vector is not None:
+ if isinstance(__Vector, str):
+ __Vector = __Vector.replace(";"," ").replace(","," ").split()
self.__is_vector = True
- self.__C = numpy.abs( numpy.array( numpy.ravel( numpy.matrix(__Vector, float ) ) ) )
+ self.__C = numpy.abs( numpy.array( numpy.ravel( __Vector ), dtype=float ) )
self.shape = (self.__C.size,self.__C.size)
self.size = self.__C.size**2
elif __Matrix is not None:
elif __Object is not None:
self.__is_object = True
self.__C = __Object
- for at in ("getT","getI","diag","trace","__add__","__sub__","__neg__","__mul__","__rmul__"):
+ for at in ("getT","getI","diag","trace","__add__","__sub__","__neg__","__matmul__","__mul__","__rmatmul__","__rmul__"):
if not hasattr(self.__C,at):
raise ValueError("The matrix given for %s as an object has no attribute \"%s\". Please check your object input."%(self.__name,at))
if hasattr(self.__C,"shape"):
def getI(self):
"Inversion"
if self.ismatrix():
- return Covariance(self.__name+"I", asCovariance = self.__C.I )
+ return Covariance(self.__name+"I", asCovariance = numpy.linalg.inv(self.__C) )
elif self.isvector():
return Covariance(self.__name+"I", asEyeByVector = 1. / self.__C )
elif self.isscalar():
return Covariance(self.__name+"I", asEyeByScalar = 1. / self.__C )
- elif self.isobject():
+ elif self.isobject() and hasattr(self.__C,"getI"):
return Covariance(self.__name+"I", asCovObject = self.__C.getI() )
else:
return None # Indispensable
return Covariance(self.__name+"T", asEyeByVector = self.__C )
elif self.isscalar():
return Covariance(self.__name+"T", asEyeByScalar = self.__C )
- elif self.isobject():
+ elif self.isobject() and hasattr(self.__C,"getT"):
return Covariance(self.__name+"T", asCovObject = self.__C.getT() )
+ else:
+ raise AttributeError("the %s covariance matrix has no getT attribute."%(self.__name,))
def cholesky(self):
"Décomposition de Cholesky"
return Covariance(self.__name+"C", asEyeByScalar = numpy.sqrt( self.__C ) )
elif self.isobject() and hasattr(self.__C,"cholesky"):
return Covariance(self.__name+"C", asCovObject = self.__C.cholesky() )
+ else:
+ raise AttributeError("the %s covariance matrix has no cholesky attribute."%(self.__name,))
def choleskyI(self):
"Inversion de la décomposition de Cholesky"
if self.ismatrix():
- return Covariance(self.__name+"H", asCovariance = numpy.linalg.cholesky(self.__C).I )
+ return Covariance(self.__name+"H", asCovariance = numpy.linalg.inv(numpy.linalg.cholesky(self.__C)) )
elif self.isvector():
return Covariance(self.__name+"H", asEyeByVector = 1.0 / numpy.sqrt( self.__C ) )
elif self.isscalar():
return Covariance(self.__name+"H", asEyeByScalar = 1.0 / numpy.sqrt( self.__C ) )
elif self.isobject() and hasattr(self.__C,"choleskyI"):
return Covariance(self.__name+"H", asCovObject = self.__C.choleskyI() )
+ else:
+ raise AttributeError("the %s covariance matrix has no choleskyI attribute."%(self.__name,))
+
+ def sqrtm(self):
+ "Racine carrée matricielle"
+ if self.ismatrix():
+ import scipy
+ return Covariance(self.__name+"C", asCovariance = numpy.real(scipy.linalg.sqrtm(self.__C)) )
+ elif self.isvector():
+ return Covariance(self.__name+"C", asEyeByVector = numpy.sqrt( self.__C ) )
+ elif self.isscalar():
+ return Covariance(self.__name+"C", asEyeByScalar = numpy.sqrt( self.__C ) )
+ elif self.isobject() and hasattr(self.__C,"sqrtm"):
+ return Covariance(self.__name+"C", asCovObject = self.__C.sqrtm() )
+ else:
+ raise AttributeError("the %s covariance matrix has no sqrtm attribute."%(self.__name,))
+
+ def sqrtmI(self):
+ "Inversion de la racine carrée matricielle"
+ if self.ismatrix():
+ import scipy
+ return Covariance(self.__name+"H", asCovariance = numpy.linalg.inv(numpy.real(scipy.linalg.sqrtm(self.__C))) )
+ elif self.isvector():
+ return Covariance(self.__name+"H", asEyeByVector = 1.0 / numpy.sqrt( self.__C ) )
+ elif self.isscalar():
+ return Covariance(self.__name+"H", asEyeByScalar = 1.0 / numpy.sqrt( self.__C ) )
+ elif self.isobject() and hasattr(self.__C,"sqrtmI"):
+ return Covariance(self.__name+"H", asCovObject = self.__C.sqrtmI() )
+ else:
+ raise AttributeError("the %s covariance matrix has no sqrtmI attribute."%(self.__name,))
def diag(self, msize=None):
"Diagonale de la matrice"
raise ValueError("the size of the %s covariance matrix has to be given in case of definition as a scalar over the diagonal."%(self.__name,))
else:
return self.__C * numpy.ones(int(msize))
- elif self.isobject():
+ elif self.isobject() and hasattr(self.__C,"diag"):
return self.__C.diag()
-
- def asfullmatrix(self, msize=None):
- "Matrice pleine"
- if self.ismatrix():
- return self.__C
- elif self.isvector():
- return numpy.matrix( numpy.diag(self.__C), float )
- elif self.isscalar():
- if msize is None:
- raise ValueError("the size of the %s covariance matrix has to be given in case of definition as a scalar over the diagonal."%(self.__name,))
- else:
- return numpy.matrix( self.__C * numpy.eye(int(msize)), float )
- elif self.isobject() and hasattr(self.__C,"asfullmatrix"):
- return self.__C.asfullmatrix()
+ else:
+ raise AttributeError("the %s covariance matrix has no diag attribute."%(self.__name,))
def trace(self, msize=None):
"Trace de la matrice"
return self.__C * int(msize)
elif self.isobject():
return self.__C.trace()
+ else:
+ raise AttributeError("the %s covariance matrix has no trace attribute."%(self.__name,))
+
+ def asfullmatrix(self, msize=None):
+ "Matrice pleine"
+ if self.ismatrix():
+ return numpy.asarray(self.__C)
+ elif self.isvector():
+ return numpy.asarray( numpy.diag(self.__C), float )
+ elif self.isscalar():
+ if msize is None:
+ raise ValueError("the size of the %s covariance matrix has to be given in case of definition as a scalar over the diagonal."%(self.__name,))
+ else:
+ return numpy.asarray( self.__C * numpy.eye(int(msize)), float )
+ elif self.isobject() and hasattr(self.__C,"asfullmatrix"):
+ return self.__C.asfullmatrix()
+ else:
+ raise AttributeError("the %s covariance matrix has no asfullmatrix attribute."%(self.__name,))
+
+ def assparsematrix(self):
+ "Valeur sparse"
+ return self.__C
def getO(self):
return self
"x.__neg__() <==> -x"
return - self.__C
+ def __matmul__(self, other):
+ "x.__mul__(y) <==> x@y"
+ if self.ismatrix() and isinstance(other, (int, float)):
+ return numpy.asarray(self.__C) * other
+ elif self.ismatrix() and isinstance(other, (list, numpy.matrix, numpy.ndarray, tuple)):
+ if numpy.ravel(other).size == self.shape[1]: # Vecteur
+ return numpy.ravel(self.__C @ numpy.ravel(other))
+ elif numpy.asarray(other).shape[0] == self.shape[1]: # Matrice
+ return numpy.asarray(self.__C) @ numpy.asarray(other)
+ else:
+ raise ValueError("operands could not be broadcast together with shapes %s %s in %s matrix"%(self.shape,numpy.asarray(other).shape,self.__name))
+ elif self.isvector() and isinstance(other, (list, numpy.matrix, numpy.ndarray, tuple)):
+ if numpy.ravel(other).size == self.shape[1]: # Vecteur
+ return numpy.ravel(self.__C) * numpy.ravel(other)
+ elif numpy.asarray(other).shape[0] == self.shape[1]: # Matrice
+ return numpy.ravel(self.__C).reshape((-1,1)) * numpy.asarray(other)
+ else:
+ raise ValueError("operands could not be broadcast together with shapes %s %s in %s matrix"%(self.shape,numpy.ravel(other).shape,self.__name))
+ elif self.isscalar() and isinstance(other,numpy.matrix):
+ return numpy.asarray(self.__C * other)
+ elif self.isscalar() and isinstance(other, (list, numpy.ndarray, tuple)):
+ if len(numpy.asarray(other).shape) == 1 or numpy.asarray(other).shape[1] == 1 or numpy.asarray(other).shape[0] == 1:
+ return self.__C * numpy.ravel(other)
+ else:
+ return self.__C * numpy.asarray(other)
+ elif self.isobject():
+ return self.__C.__matmul__(other)
+ else:
+ raise NotImplementedError("%s covariance matrix __matmul__ method not available for %s type!"%(self.__name,type(other)))
+
def __mul__(self, other):
"x.__mul__(y) <==> x*y"
if self.ismatrix() and isinstance(other, (int, numpy.matrix, float)):
else:
raise NotImplementedError("%s covariance matrix __mul__ method not available for %s type!"%(self.__name,type(other)))
+ def __rmatmul__(self, other):
+ "x.__rmul__(y) <==> y@x"
+ if self.ismatrix() and isinstance(other, (int, numpy.matrix, float)):
+ return other * self.__C
+ elif self.ismatrix() and isinstance(other, (list, numpy.ndarray, tuple)):
+ if numpy.ravel(other).size == self.shape[1]: # Vecteur
+ return numpy.asmatrix(numpy.ravel(other)) * self.__C
+ elif numpy.asmatrix(other).shape[0] == self.shape[1]: # Matrice
+ return numpy.asmatrix(other) * self.__C
+ else:
+ raise ValueError("operands could not be broadcast together with shapes %s %s in %s matrix"%(numpy.asmatrix(other).shape,self.shape,self.__name))
+ elif self.isvector() and isinstance(other,numpy.matrix):
+ if numpy.ravel(other).size == self.shape[0]: # Vecteur
+ return numpy.asmatrix(numpy.ravel(other) * self.__C)
+ elif numpy.asmatrix(other).shape[1] == self.shape[0]: # Matrice
+ return numpy.asmatrix(numpy.array(other) * self.__C)
+ else:
+ raise ValueError("operands could not be broadcast together with shapes %s %s in %s matrix"%(numpy.ravel(other).shape,self.shape,self.__name))
+ elif self.isscalar() and isinstance(other,numpy.matrix):
+ return other * self.__C
+ elif self.isobject():
+ return self.__C.__rmatmul__(other)
+ else:
+ raise NotImplementedError("%s covariance matrix __rmatmul__ method not available for %s type!"%(self.__name,type(other)))
+
def __rmul__(self, other):
"x.__rmul__(y) <==> y*x"
if self.ismatrix() and isinstance(other, (int, numpy.matrix, float)):
#
# Calculs effectifs
if __mpEnabled:
- _jobs = []
- if _extraArguments is None:
- _jobs = __xserie
- elif _extraArguments is not None and isinstance(_extraArguments, (list, tuple, map)):
- for __xvalue in __xserie:
- _jobs.append( [__xvalue, ] + list(_extraArguments) )
- else:
- raise TypeError("MultiFonction extra arguments unkown input type: %s"%(type(_extraArguments),))
+ _jobs = __xserie
# logging.debug("MULTF Internal multiprocessing calculations begin : evaluation of %i point(s)"%(len(_jobs),))
import multiprocessing
with multiprocessing.Pool(__mpWorkers) as pool: