src/daComposant/daAlgorithms/Atoms/ecwnlls.py

   1 # -*- coding: utf-8 -*-
   2 #
   3 # Copyright (C) 2008-2022 EDF R&D
   4 #
   5 # This library is free software; you can redistribute it and/or
   6 # modify it under the terms of the GNU Lesser General Public
   7 # License as published by the Free Software Foundation; either
   8 # version 2.1 of the License.
   9 #
  10 # This library is distributed in the hope that it will be useful,
  11 # but WITHOUT ANY WARRANTY; without even the implied warranty of
  12 # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  13 # Lesser General Public License for more details.
  14 #
  15 # You should have received a copy of the GNU Lesser General Public
  16 # License along with this library; if not, write to the Free Software
  17 # Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307 USA
  18 #
  19 # See http://www.salome-platform.org/ or email : webmaster.salome@opencascade.com
  20 #
  21 # Author: Jean-Philippe Argaud, jean-philippe.argaud@edf.fr, EDF R&D
  22
  23 __doc__ = """
  24     Non Linear Least Squares
  25 """
  26 __author__ = "Jean-Philippe ARGAUD"
  27
  28 import numpy, scipy, scipy.optimize, scipy.version
  29
  30 # ==============================================================================
  31 def ecwnlls(selfA, Xb, Y, U, HO, CM, R, B, __storeState = False):
  32     """
  33     Correction
  34     """
  35     #
  36     # Initialisations
  37     # ---------------
  38     Hm = HO["Direct"].appliedTo
  39     Ha = HO["Adjoint"].appliedInXTo
  40     #
  41     if HO["AppliedInX"] is not None and "HXb" in HO["AppliedInX"]:
  42         HXb = numpy.asarray(Hm( Xb, HO["AppliedInX"]["HXb"] ))
  43     else:
  44         HXb = numpy.asarray(Hm( Xb ))
  45     HXb = HXb.reshape((-1,1))
  46     if Y.size != HXb.size:
  47         raise ValueError("The size %i of observations Y and %i of observed calculation H(X) are different, they have to be identical."%(Y.size,HXb.size))
  48     if max(Y.shape) != max(HXb.shape):
  49         raise ValueError("The shapes %s of observations Y and %s of observed calculation H(X) are different, they have to be identical."%(Y.shape,HXb.shape))
  50     #
  51     RI = R.getI()
  52     if selfA._parameters["Minimizer"] == "LM":
  53         RdemiI = R.choleskyI()
  54     #
  55     Xini = selfA._parameters["InitializationPoint"]
  56     #
  57     # Définition de la fonction-coût
  58     # ------------------------------
  59     def CostFunction(x):
  60         _X  = numpy.asarray(x).reshape((-1,1))
  61         if selfA._parameters["StoreInternalVariables"] or \
  62             selfA._toStore("CurrentState") or \
  63             selfA._toStore("CurrentOptimum"):
  64             selfA.StoredVariables["CurrentState"].store( _X )
  65         _HX = numpy.asarray(Hm( _X )).reshape((-1,1))
  66         _Innovation = Y - _HX
  67         if selfA._toStore("SimulatedObservationAtCurrentState") or \
  68             selfA._toStore("SimulatedObservationAtCurrentOptimum"):
  69             selfA.StoredVariables["SimulatedObservationAtCurrentState"].store( _HX )
  70         if selfA._toStore("InnovationAtCurrentState"):
  71             selfA.StoredVariables["InnovationAtCurrentState"].store( _Innovation )
  72         #
  73         Jb  = 0.
  74         Jo  = float( 0.5 * _Innovation.T * (RI * _Innovation) )
  75         J   = Jb + Jo
  76         #
  77         selfA.StoredVariables["CurrentIterationNumber"].store( len(selfA.StoredVariables["CostFunctionJ"]) )
  78         selfA.StoredVariables["CostFunctionJb"].store( Jb )
  79         selfA.StoredVariables["CostFunctionJo"].store( Jo )
  80         selfA.StoredVariables["CostFunctionJ" ].store( J )
  81         if selfA._toStore("IndexOfOptimum") or \
  82             selfA._toStore("CurrentOptimum") or \
  83             selfA._toStore("CostFunctionJAtCurrentOptimum") or \
  84             selfA._toStore("CostFunctionJbAtCurrentOptimum") or \
  85             selfA._toStore("CostFunctionJoAtCurrentOptimum") or \
  86             selfA._toStore("SimulatedObservationAtCurrentOptimum"):
  87             IndexMin = numpy.argmin( selfA.StoredVariables["CostFunctionJ"][nbPreviousSteps:] ) + nbPreviousSteps
  88         if selfA._toStore("IndexOfOptimum"):
  89             selfA.StoredVariables["IndexOfOptimum"].store( IndexMin )
  90         if selfA._toStore("CurrentOptimum"):
  91             selfA.StoredVariables["CurrentOptimum"].store( selfA.StoredVariables["CurrentState"][IndexMin] )
  92         if selfA._toStore("SimulatedObservationAtCurrentOptimum"):
  93             selfA.StoredVariables["SimulatedObservationAtCurrentOptimum"].store( selfA.StoredVariables["SimulatedObservationAtCurrentState"][IndexMin] )
  94         if selfA._toStore("CostFunctionJbAtCurrentOptimum"):
  95             selfA.StoredVariables["CostFunctionJbAtCurrentOptimum"].store( selfA.StoredVariables["CostFunctionJb"][IndexMin] )
  96         if selfA._toStore("CostFunctionJoAtCurrentOptimum"):
  97             selfA.StoredVariables["CostFunctionJoAtCurrentOptimum"].store( selfA.StoredVariables["CostFunctionJo"][IndexMin] )
  98         if selfA._toStore("CostFunctionJAtCurrentOptimum"):
  99             selfA.StoredVariables["CostFunctionJAtCurrentOptimum" ].store( selfA.StoredVariables["CostFunctionJ" ][IndexMin] )
 100         return J
 101     #
 102     def GradientOfCostFunction(x):
 103         _X      = numpy.asarray(x).reshape((-1,1))
 104         _HX     = numpy.asarray(Hm( _X )).reshape((-1,1))
 105         GradJb  = 0.
 106         GradJo  = - Ha( (_X, RI * (Y - _HX)) )
 107         GradJ   = numpy.ravel( GradJb ) + numpy.ravel( GradJo )
 108         return GradJ
 109     #
 110     def CostFunctionLM(x):
 111         _X  = numpy.ravel( x ).reshape((-1,1))
 112         _HX = Hm( _X ).reshape((-1,1))
 113         _Innovation = Y - _HX
 114         Jb  = 0.
 115         Jo  = float( 0.5 * _Innovation.T * (RI * _Innovation) )
 116         J   = Jb + Jo
 117         if selfA._parameters["StoreInternalVariables"] or \
 118             selfA._toStore("CurrentState"):
 119             selfA.StoredVariables["CurrentState"].store( _X )
 120         selfA.StoredVariables["CostFunctionJb"].store( Jb )
 121         selfA.StoredVariables["CostFunctionJo"].store( Jo )
 122         selfA.StoredVariables["CostFunctionJ" ].store( J )
 123         #
 124         return numpy.ravel( RdemiI*_Innovation )
 125     #
 126     def GradientOfCostFunctionLM(x):
 127         _X      = x.reshape((-1,1))
 128         return - RdemiI*HO["Tangent"].asMatrix( _X )
 129     #
 130     # Minimisation de la fonctionnelle
 131     # --------------------------------
 132     nbPreviousSteps = selfA.StoredVariables["CostFunctionJ"].stepnumber()
 133     #
 134     if selfA._parameters["Minimizer"] == "LBFGSB":
 135         if "0.19" <= scipy.version.version <= "1.4.99":
 136             import daAlgorithms.Atoms.lbfgsb14hlt as optimiseur
 137         elif "1.5.0" <= scipy.version.version <= "1.7.99":
 138             import daAlgorithms.Atoms.lbfgsb17hlt as optimiseur
 139         elif "1.8.0" <= scipy.version.version <= "1.8.99":
 140             import daAlgorithms.Atoms.lbfgsb18hlt as optimiseur
 141         elif "1.9.0" <= scipy.version.version <= "1.9.99":
 142             import daAlgorithms.Atoms.lbfgsb19hlt as optimiseur
 143         else:
 144             import scipy.optimize as optimiseur
 145         Minimum, J_optimal, Informations = optimiseur.fmin_l_bfgs_b(
 146             func        = CostFunction,
 147             x0          = Xini,
 148             fprime      = GradientOfCostFunction,
 149             args        = (),
 150             bounds      = selfA._parameters["Bounds"],
 151             maxfun      = selfA._parameters["MaximumNumberOfIterations"]-1,
 152             factr       = selfA._parameters["CostDecrementTolerance"]*1.e14,
 153             pgtol       = selfA._parameters["ProjectedGradientTolerance"],
 154             iprint      = selfA._parameters["optiprint"],
 155             )
 156         # nfeval = Informations['funcalls']
 157         # rc     = Informations['warnflag']
 158     elif selfA._parameters["Minimizer"] == "TNC":
 159         Minimum, nfeval, rc = scipy.optimize.fmin_tnc(
 160             func        = CostFunction,
 161             x0          = Xini,
 162             fprime      = GradientOfCostFunction,
 163             args        = (),
 164             bounds      = selfA._parameters["Bounds"],
 165             maxfun      = selfA._parameters["MaximumNumberOfIterations"],
 166             pgtol       = selfA._parameters["ProjectedGradientTolerance"],
 167             ftol        = selfA._parameters["CostDecrementTolerance"],
 168             messages    = selfA._parameters["optmessages"],
 169             )
 170     elif selfA._parameters["Minimizer"] == "CG":
 171         Minimum, fopt, nfeval, grad_calls, rc = scipy.optimize.fmin_cg(
 172             f           = CostFunction,
 173             x0          = Xini,
 174             fprime      = GradientOfCostFunction,
 175             args        = (),
 176             maxiter     = selfA._parameters["MaximumNumberOfIterations"],
 177             gtol        = selfA._parameters["GradientNormTolerance"],
 178             disp        = selfA._parameters["optdisp"],
 179             full_output = True,
 180             )
 181     elif selfA._parameters["Minimizer"] == "NCG":
 182         Minimum, fopt, nfeval, grad_calls, hcalls, rc = scipy.optimize.fmin_ncg(
 183             f           = CostFunction,
 184             x0          = Xini,
 185             fprime      = GradientOfCostFunction,
 186             args        = (),
 187             maxiter     = selfA._parameters["MaximumNumberOfIterations"],
 188             avextol     = selfA._parameters["CostDecrementTolerance"],
 189             disp        = selfA._parameters["optdisp"],
 190             full_output = True,
 191             )
 192     elif selfA._parameters["Minimizer"] == "BFGS":
 193         Minimum, fopt, gopt, Hopt, nfeval, grad_calls, rc = scipy.optimize.fmin_bfgs(
 194             f           = CostFunction,
 195             x0          = Xini,
 196             fprime      = GradientOfCostFunction,
 197             args        = (),
 198             maxiter     = selfA._parameters["MaximumNumberOfIterations"],
 199             gtol        = selfA._parameters["GradientNormTolerance"],
 200             disp        = selfA._parameters["optdisp"],
 201             full_output = True,
 202             )
 203     elif selfA._parameters["Minimizer"] == "LM":
 204         Minimum, cov_x, infodict, mesg, rc = scipy.optimize.leastsq(
 205             func        = CostFunctionLM,
 206             x0          = Xini,
 207             Dfun        = GradientOfCostFunctionLM,
 208             args        = (),
 209             ftol        = selfA._parameters["CostDecrementTolerance"],
 210             maxfev      = selfA._parameters["MaximumNumberOfIterations"],
 211             gtol        = selfA._parameters["GradientNormTolerance"],
 212             full_output = True,
 213             )
 214         # nfeval = infodict['nfev']
 215     else:
 216         raise ValueError("Error in minimizer name: %s is unkown"%selfA._parameters["Minimizer"])
 217     #
 218     IndexMin = numpy.argmin( selfA.StoredVariables["CostFunctionJ"][nbPreviousSteps:] ) + nbPreviousSteps
 219     #
 220     # Correction pour pallier a un bug de TNC sur le retour du Minimum
 221     # ----------------------------------------------------------------
 222     if selfA._parameters["StoreInternalVariables"] or selfA._toStore("CurrentState"):
 223         Minimum = selfA.StoredVariables["CurrentState"][IndexMin]
 224     #
 225     Xa = Minimum
 226     if __storeState: selfA._setInternalState("Xn", Xa)
 227     #--------------------------
 228     #
 229     selfA.StoredVariables["Analysis"].store( Xa )
 230     #
 231     if selfA._toStore("OMA") or \
 232         selfA._toStore("InnovationAtCurrentAnalysis") or \
 233         selfA._toStore("SimulatedObservationAtOptimum"):
 234         if selfA._toStore("SimulatedObservationAtCurrentState"):
 235             HXa = selfA.StoredVariables["SimulatedObservationAtCurrentState"][IndexMin]
 236         elif selfA._toStore("SimulatedObservationAtCurrentOptimum"):
 237             HXa = selfA.StoredVariables["SimulatedObservationAtCurrentOptimum"][-1]
 238         else:
 239             HXa = Hm( Xa )
 240         oma = Y - HXa.reshape((-1,1))
 241     #
 242     # Calculs et/ou stockages supplémentaires
 243     # ---------------------------------------
 244     if selfA._toStore("Innovation") or \
 245         selfA._toStore("OMB"):
 246         Innovation  = Y - HXb
 247     if selfA._toStore("Innovation"):
 248         selfA.StoredVariables["Innovation"].store( Innovation )
 249     if selfA._toStore("BMA"):
 250         selfA.StoredVariables["BMA"].store( numpy.ravel(Xb) - numpy.ravel(Xa) )
 251     if selfA._toStore("OMA"):
 252         selfA.StoredVariables["OMA"].store( oma )
 253     if selfA._toStore("InnovationAtCurrentAnalysis"):
 254         selfA.StoredVariables["InnovationAtCurrentAnalysis"].store( oma )
 255     if selfA._toStore("OMB"):
 256         selfA.StoredVariables["OMB"].store( Innovation )
 257     if selfA._toStore("SimulatedObservationAtBackground"):
 258         selfA.StoredVariables["SimulatedObservationAtBackground"].store( HXb )
 259     if selfA._toStore("SimulatedObservationAtOptimum"):
 260         selfA.StoredVariables["SimulatedObservationAtOptimum"].store( HXa )
 261     #
 262     return 0
 263
 264 # ==============================================================================
 265 if __name__ == "__main__":
 266     print('\n AUTODIAGNOSTIC\n')