src/daComposant/daAlgorithms/Atoms/ecwnlls.py

   1 # -*- coding: utf-8 -*-
   2 #
   3 # Copyright (C) 2008-2024 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 from daCore.PlatformInfo import vt, vfloat
  30
  31 # ==============================================================================
  32 def ecwnlls(selfA, Xb, Y, U, HO, CM, R, B, __storeState = False):
  33     """
  34     Correction
  35     """
  36     #
  37     # Initialisations
  38     # ---------------
  39     Hm = HO["Direct"].appliedTo
  40     Ha = HO["Adjoint"].appliedInXTo
  41     #
  42     if HO["AppliedInX"] is not None and "HXb" in HO["AppliedInX"]:
  43         HXb = numpy.asarray(Hm( Xb, HO["AppliedInX"]["HXb"] ))
  44     else:
  45         HXb = numpy.asarray(Hm( Xb ))
  46     HXb = HXb.reshape((-1, 1))
  47     if Y.size != HXb.size:
  48         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))  # noqa: E501
  49     if max(Y.shape) != max(HXb.shape):
  50         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))  # noqa: E501
  51     #
  52     RI = R.getI()
  53     if selfA._parameters["Minimizer"] == "LM":
  54         RdemiI = R.choleskyI()
  55     #
  56     Xini = selfA._parameters["InitializationPoint"]
  57     #
  58     # Définition de la fonction-coût
  59     # ------------------------------
  60
  61     def CostFunction(x):
  62         _X  = numpy.asarray(x).reshape((-1, 1))
  63         if selfA._parameters["StoreInternalVariables"] or \
  64                 selfA._toStore("CurrentState") or \
  65                 selfA._toStore("CurrentOptimum"):
  66             selfA.StoredVariables["CurrentState"].store( _X )
  67         _HX = numpy.asarray(Hm( _X )).reshape((-1, 1))
  68         _Innovation = Y - _HX
  69         if selfA._toStore("SimulatedObservationAtCurrentState") or \
  70                 selfA._toStore("SimulatedObservationAtCurrentOptimum"):
  71             selfA.StoredVariables["SimulatedObservationAtCurrentState"].store( _HX )
  72         if selfA._toStore("InnovationAtCurrentState"):
  73             selfA.StoredVariables["InnovationAtCurrentState"].store( _Innovation )
  74         #
  75         Jb  = 0.
  76         Jo  = vfloat( 0.5 * _Innovation.T * (RI * _Innovation) )
  77         J   = Jb + Jo
  78         #
  79         selfA.StoredVariables["CurrentIterationNumber"].store( len(selfA.StoredVariables["CostFunctionJ"]) )
  80         selfA.StoredVariables["CostFunctionJb"].store( Jb )
  81         selfA.StoredVariables["CostFunctionJo"].store( Jo )
  82         selfA.StoredVariables["CostFunctionJ" ].store( J )
  83         if selfA._toStore("IndexOfOptimum") or \
  84                 selfA._toStore("CurrentOptimum") or \
  85                 selfA._toStore("CostFunctionJAtCurrentOptimum") or \
  86                 selfA._toStore("CostFunctionJbAtCurrentOptimum") or \
  87                 selfA._toStore("CostFunctionJoAtCurrentOptimum") or \
  88                 selfA._toStore("SimulatedObservationAtCurrentOptimum"):
  89             IndexMin = numpy.argmin( selfA.StoredVariables["CostFunctionJ"][nbPreviousSteps:] ) + nbPreviousSteps
  90         if selfA._toStore("IndexOfOptimum"):
  91             selfA.StoredVariables["IndexOfOptimum"].store( IndexMin )
  92         if selfA._toStore("CurrentOptimum"):
  93             selfA.StoredVariables["CurrentOptimum"].store( selfA.StoredVariables["CurrentState"][IndexMin] )
  94         if selfA._toStore("SimulatedObservationAtCurrentOptimum"):
  95             selfA.StoredVariables["SimulatedObservationAtCurrentOptimum"].store( selfA.StoredVariables["SimulatedObservationAtCurrentState"][IndexMin] )  # noqa: E501
  96         if selfA._toStore("CostFunctionJbAtCurrentOptimum"):
  97             selfA.StoredVariables["CostFunctionJbAtCurrentOptimum"].store( selfA.StoredVariables["CostFunctionJb"][IndexMin] )  # noqa: E501
  98         if selfA._toStore("CostFunctionJoAtCurrentOptimum"):
  99             selfA.StoredVariables["CostFunctionJoAtCurrentOptimum"].store( selfA.StoredVariables["CostFunctionJo"][IndexMin] )  # noqa: E501
 100         if selfA._toStore("CostFunctionJAtCurrentOptimum"):
 101             selfA.StoredVariables["CostFunctionJAtCurrentOptimum" ].store( selfA.StoredVariables["CostFunctionJ" ][IndexMin] )  # noqa: E501
 102         return J
 103
 104     def GradientOfCostFunction(x):
 105         _X      = numpy.asarray(x).reshape((-1, 1))
 106         _HX     = numpy.asarray(Hm( _X )).reshape((-1, 1))
 107         GradJb  = 0.
 108         GradJo  = - Ha( (_X, RI * (Y - _HX)) )
 109         GradJ   = numpy.ravel( GradJb ) + numpy.ravel( GradJo )
 110         return GradJ
 111
 112     def CostFunctionLM(x):
 113         _X  = numpy.ravel( x ).reshape((-1, 1))
 114         _HX = Hm( _X ).reshape((-1, 1))
 115         _Innovation = Y - _HX
 116         Jb  = 0.
 117         Jo  = vfloat( 0.5 * _Innovation.T * (RI * _Innovation) )
 118         J   = Jb + Jo
 119         if selfA._parameters["StoreInternalVariables"] or \
 120                 selfA._toStore("CurrentState"):
 121             selfA.StoredVariables["CurrentState"].store( _X )
 122         selfA.StoredVariables["CostFunctionJb"].store( Jb )
 123         selfA.StoredVariables["CostFunctionJo"].store( Jo )
 124         selfA.StoredVariables["CostFunctionJ" ].store( J )
 125         #
 126         return numpy.ravel( RdemiI * _Innovation )
 127
 128     def GradientOfCostFunctionLM(x):
 129         _X      = x.reshape((-1, 1))
 130         return - RdemiI * HO["Tangent"].asMatrix( _X )
 131     #
 132     # Minimisation de la fonctionnelle
 133     # --------------------------------
 134     nbPreviousSteps = selfA.StoredVariables["CostFunctionJ"].stepnumber()
 135     #
 136     if selfA._parameters["Minimizer"] == "LBFGSB":
 137         if vt("0.19")  <= vt(scipy.version.version) <= vt("1.4.99"):
 138             import daAlgorithms.Atoms.lbfgsb14hlt as optimiseur
 139         elif vt("1.5.0") <= vt(scipy.version.version) <= vt("1.7.99"):
 140             import daAlgorithms.Atoms.lbfgsb17hlt as optimiseur
 141         elif vt("1.8.0") <= vt(scipy.version.version) <= vt("1.8.99"):
 142             import daAlgorithms.Atoms.lbfgsb18hlt as optimiseur
 143         elif vt("1.9.0") <= vt(scipy.version.version) <= vt("1.10.99"):
 144             import daAlgorithms.Atoms.lbfgsb19hlt as optimiseur
 145         elif vt("1.11.0") <= vt(scipy.version.version) <= vt("1.11.99"):
 146             import daAlgorithms.Atoms.lbfgsb111hlt as optimiseur
 147         elif vt("1.12.0") <= vt(scipy.version.version) <= vt("1.12.99"):
 148             import daAlgorithms.Atoms.lbfgsb112hlt as optimiseur
 149         elif vt("1.13.0") <= vt(scipy.version.version) <= vt("1.13.99"):
 150             import daAlgorithms.Atoms.lbfgsb113hlt as optimiseur
 151         elif vt("1.14.0") <= vt(scipy.version.version) <= vt("1.14.99"):
 152             import daAlgorithms.Atoms.lbfgsb114hlt as optimiseur
 153         else:
 154             import scipy.optimize as optimiseur
 155         Minimum, J_optimal, Informations = optimiseur.fmin_l_bfgs_b(
 156             func        = CostFunction,
 157             x0          = Xini,
 158             fprime      = GradientOfCostFunction,
 159             args        = (),
 160             bounds      = selfA._parameters["Bounds"],
 161             maxfun      = selfA._parameters["MaximumNumberOfIterations"] - 1,
 162             factr       = selfA._parameters["CostDecrementTolerance"] * 1.e14,
 163             pgtol       = selfA._parameters["ProjectedGradientTolerance"],
 164             iprint      = selfA._parameters["optiprint"],
 165         )
 166         # nfeval = Informations['funcalls']
 167         # rc     = Informations['warnflag']
 168     elif selfA._parameters["Minimizer"] == "TNC":
 169         Minimum, nfeval, rc = scipy.optimize.fmin_tnc(
 170             func        = CostFunction,
 171             x0          = Xini,
 172             fprime      = GradientOfCostFunction,
 173             args        = (),
 174             bounds      = selfA._parameters["Bounds"],
 175             maxfun      = selfA._parameters["MaximumNumberOfIterations"],
 176             pgtol       = selfA._parameters["ProjectedGradientTolerance"],
 177             ftol        = selfA._parameters["CostDecrementTolerance"],
 178             messages    = selfA._parameters["optmessages"],
 179         )
 180     elif selfA._parameters["Minimizer"] == "CG":
 181         Minimum, fopt, nfeval, grad_calls, rc = scipy.optimize.fmin_cg(
 182             f           = CostFunction,
 183             x0          = Xini,
 184             fprime      = GradientOfCostFunction,
 185             args        = (),
 186             maxiter     = selfA._parameters["MaximumNumberOfIterations"],
 187             gtol        = selfA._parameters["GradientNormTolerance"],
 188             disp        = selfA._parameters["optdisp"],
 189             full_output = True,
 190         )
 191     elif selfA._parameters["Minimizer"] == "NCG":
 192         Minimum, fopt, nfeval, grad_calls, hcalls, rc = scipy.optimize.fmin_ncg(
 193             f           = CostFunction,
 194             x0          = Xini,
 195             fprime      = GradientOfCostFunction,
 196             args        = (),
 197             maxiter     = selfA._parameters["MaximumNumberOfIterations"],
 198             avextol     = selfA._parameters["CostDecrementTolerance"],
 199             disp        = selfA._parameters["optdisp"],
 200             full_output = True,
 201         )
 202     elif selfA._parameters["Minimizer"] == "BFGS":
 203         Minimum, fopt, gopt, Hopt, nfeval, grad_calls, rc = scipy.optimize.fmin_bfgs(
 204             f           = CostFunction,
 205             x0          = Xini,
 206             fprime      = GradientOfCostFunction,
 207             args        = (),
 208             maxiter     = selfA._parameters["MaximumNumberOfIterations"],
 209             gtol        = selfA._parameters["GradientNormTolerance"],
 210             disp        = selfA._parameters["optdisp"],
 211             full_output = True,
 212         )
 213     elif selfA._parameters["Minimizer"] == "LM":
 214         Minimum, cov_x, infodict, mesg, rc = scipy.optimize.leastsq(
 215             func        = CostFunctionLM,
 216             x0          = Xini,
 217             Dfun        = GradientOfCostFunctionLM,
 218             args        = (),
 219             ftol        = selfA._parameters["CostDecrementTolerance"],
 220             maxfev      = selfA._parameters["MaximumNumberOfIterations"],
 221             gtol        = selfA._parameters["GradientNormTolerance"],
 222             full_output = True,
 223         )
 224         # nfeval = infodict['nfev']
 225     else:
 226         raise ValueError("Error in minimizer name: %s is unkown"%selfA._parameters["Minimizer"])
 227     #
 228     IndexMin = numpy.argmin( selfA.StoredVariables["CostFunctionJ"][nbPreviousSteps:] ) + nbPreviousSteps
 229     #
 230     # Correction pour pallier a un bug de TNC sur le retour du Minimum
 231     # ----------------------------------------------------------------
 232     if selfA._parameters["StoreInternalVariables"] or selfA._toStore("CurrentState"):
 233         Minimum = selfA.StoredVariables["CurrentState"][IndexMin]
 234     #
 235     Xa = Minimum
 236     if __storeState:
 237         selfA._setInternalState("Xn", Xa)
 238     # --------------------------
 239     #
 240     selfA.StoredVariables["Analysis"].store( Xa )
 241     #
 242     if selfA._toStore("OMA") or \
 243             selfA._toStore("InnovationAtCurrentAnalysis") or \
 244             selfA._toStore("SimulatedObservationAtOptimum"):
 245         if selfA._toStore("SimulatedObservationAtCurrentState"):
 246             HXa = selfA.StoredVariables["SimulatedObservationAtCurrentState"][IndexMin]
 247         elif selfA._toStore("SimulatedObservationAtCurrentOptimum"):
 248             HXa = selfA.StoredVariables["SimulatedObservationAtCurrentOptimum"][-1]
 249         else:
 250             HXa = Hm( Xa )
 251         oma = Y - numpy.asarray(HXa).reshape((-1, 1))
 252     #
 253     # Calculs et/ou stockages supplémentaires
 254     # ---------------------------------------
 255     if selfA._toStore("Innovation") or \
 256             selfA._toStore("OMB"):
 257         Innovation  = Y - HXb
 258     if selfA._toStore("Innovation"):
 259         selfA.StoredVariables["Innovation"].store( Innovation )
 260     if selfA._toStore("BMA"):
 261         selfA.StoredVariables["BMA"].store( numpy.ravel(Xb) - numpy.ravel(Xa) )
 262     if selfA._toStore("OMA"):
 263         selfA.StoredVariables["OMA"].store( oma )
 264     if selfA._toStore("InnovationAtCurrentAnalysis"):
 265         selfA.StoredVariables["InnovationAtCurrentAnalysis"].store( oma )
 266     if selfA._toStore("OMB"):
 267         selfA.StoredVariables["OMB"].store( Innovation )
 268     if selfA._toStore("SimulatedObservationAtBackground"):
 269         selfA.StoredVariables["SimulatedObservationAtBackground"].store( HXb )
 270     if selfA._toStore("SimulatedObservationAtOptimum"):
 271         selfA.StoredVariables["SimulatedObservationAtOptimum"].store( HXa )
 272     #
 273     return 0
 274
 275 # ==============================================================================
 276 if __name__ == "__main__":
 277     print('\n AUTODIAGNOSTIC\n')