src/daComposant/daAlgorithms/Atoms/std3dvar.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     3DVAR
  25 """
  26 __author__ = "Jean-Philippe ARGAUD"
  27
  28 import numpy, scipy, scipy.optimize, scipy.version
  29 from daCore.NumericObjects import HessienneEstimation, QuantilesEstimations
  30 from daCore.PlatformInfo import vt, vfloat
  31
  32 # ==============================================================================
  33 def std3dvar(selfA, Xb, Y, U, HO, CM, R, B, __storeState = False):
  34     """
  35     Correction
  36     """
  37     #
  38     # Initialisations
  39     # ---------------
  40     Hm = HO["Direct"].appliedTo
  41     Ha = HO["Adjoint"].appliedInXTo
  42     #
  43     if HO["AppliedInX"] is not None and "HXb" in HO["AppliedInX"]:
  44         HXb = numpy.asarray(Hm( Xb, HO["AppliedInX"]["HXb"] ))
  45     else:
  46         HXb = numpy.asarray(Hm( Xb ))
  47     HXb = HXb.reshape((-1, 1))
  48     if Y.size != HXb.size:
  49         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
  50     if max(Y.shape) != max(HXb.shape):
  51         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
  52     #
  53     if selfA._toStore("JacobianMatrixAtBackground"):
  54         HtMb = HO["Tangent"].asMatrix(Xb)
  55         HtMb = HtMb.reshape(Y.size, Xb.size)  # ADAO & check shape
  56         selfA.StoredVariables["JacobianMatrixAtBackground"].store( HtMb )
  57     #
  58     BI = B.getI()
  59     RI = R.getI()
  60     #
  61     Xini = selfA._parameters["InitializationPoint"]
  62     #
  63     # Définition de la fonction-coût
  64     # ------------------------------
  65
  66     def CostFunction(x):
  67         _X  = numpy.asarray(x).reshape((-1, 1))
  68         if selfA._parameters["StoreInternalVariables"] or \
  69                 selfA._toStore("CurrentState") or \
  70                 selfA._toStore("CurrentOptimum"):
  71             selfA.StoredVariables["CurrentState"].store( _X )
  72         _HX = numpy.asarray(Hm( _X )).reshape((-1, 1))
  73         _Innovation = Y - _HX
  74         if selfA._toStore("SimulatedObservationAtCurrentState") or \
  75                 selfA._toStore("SimulatedObservationAtCurrentOptimum"):
  76             selfA.StoredVariables["SimulatedObservationAtCurrentState"].store( _HX )
  77         if selfA._toStore("InnovationAtCurrentState"):
  78             selfA.StoredVariables["InnovationAtCurrentState"].store( _Innovation )
  79         #
  80         Jb  = vfloat( 0.5 * (_X - Xb).T * (BI * (_X - Xb)) )
  81         Jo  = vfloat( 0.5 * _Innovation.T * (RI * _Innovation) )
  82         J   = Jb + Jo
  83         #
  84         selfA.StoredVariables["CurrentIterationNumber"].store( len(selfA.StoredVariables["CostFunctionJ"]) )
  85         selfA.StoredVariables["CostFunctionJb"].store( Jb )
  86         selfA.StoredVariables["CostFunctionJo"].store( Jo )
  87         selfA.StoredVariables["CostFunctionJ" ].store( J )
  88         if selfA._toStore("IndexOfOptimum") or \
  89                 selfA._toStore("CurrentOptimum") or \
  90                 selfA._toStore("CostFunctionJAtCurrentOptimum") or \
  91                 selfA._toStore("CostFunctionJbAtCurrentOptimum") or \
  92                 selfA._toStore("CostFunctionJoAtCurrentOptimum") or \
  93                 selfA._toStore("SimulatedObservationAtCurrentOptimum"):
  94             IndexMin = numpy.argmin( selfA.StoredVariables["CostFunctionJ"][nbPreviousSteps:] ) + nbPreviousSteps
  95         if selfA._toStore("IndexOfOptimum"):
  96             selfA.StoredVariables["IndexOfOptimum"].store( IndexMin )
  97         if selfA._toStore("CurrentOptimum"):
  98             selfA.StoredVariables["CurrentOptimum"].store( selfA.StoredVariables["CurrentState"][IndexMin] )
  99         if selfA._toStore("SimulatedObservationAtCurrentOptimum"):
 100             selfA.StoredVariables["SimulatedObservationAtCurrentOptimum"].store( selfA.StoredVariables["SimulatedObservationAtCurrentState"][IndexMin] )  # noqa: E501
 101         if selfA._toStore("CostFunctionJbAtCurrentOptimum"):
 102             selfA.StoredVariables["CostFunctionJbAtCurrentOptimum"].store( selfA.StoredVariables["CostFunctionJb"][IndexMin] )  # noqa: E501
 103         if selfA._toStore("CostFunctionJoAtCurrentOptimum"):
 104             selfA.StoredVariables["CostFunctionJoAtCurrentOptimum"].store( selfA.StoredVariables["CostFunctionJo"][IndexMin] )  # noqa: E501
 105         if selfA._toStore("CostFunctionJAtCurrentOptimum"):
 106             selfA.StoredVariables["CostFunctionJAtCurrentOptimum" ].store( selfA.StoredVariables["CostFunctionJ" ][IndexMin] )  # noqa: E501
 107         return J
 108
 109     def GradientOfCostFunction(x):
 110         _X      = numpy.asarray(x).reshape((-1, 1))
 111         _HX     = numpy.asarray(Hm( _X )).reshape((-1, 1))
 112         GradJb  = BI * (_X - Xb)
 113         GradJo  = - Ha( (_X, RI * (Y - _HX)) )
 114         GradJ   = numpy.ravel( GradJb ) + numpy.ravel( GradJo )
 115         return GradJ
 116     #
 117     # Minimisation de la fonctionnelle
 118     # --------------------------------
 119     nbPreviousSteps = selfA.StoredVariables["CostFunctionJ"].stepnumber()
 120     #
 121     if selfA._parameters["Minimizer"] == "LBFGSB":
 122         if vt("0.19")  <= vt(scipy.version.version) <= vt("1.4.99"):
 123             import daAlgorithms.Atoms.lbfgsb14hlt as optimiseur
 124         elif vt("1.5.0") <= vt(scipy.version.version) <= vt("1.7.99"):
 125             import daAlgorithms.Atoms.lbfgsb17hlt as optimiseur
 126         elif vt("1.8.0") <= vt(scipy.version.version) <= vt("1.8.99"):
 127             import daAlgorithms.Atoms.lbfgsb18hlt as optimiseur
 128         elif vt("1.9.0") <= vt(scipy.version.version) <= vt("1.10.99"):
 129             import daAlgorithms.Atoms.lbfgsb19hlt as optimiseur
 130         elif vt("1.11.0") <= vt(scipy.version.version) <= vt("1.11.99"):
 131             import daAlgorithms.Atoms.lbfgsb111hlt as optimiseur
 132         elif vt("1.12.0") <= vt(scipy.version.version) <= vt("1.12.99"):
 133             import daAlgorithms.Atoms.lbfgsb112hlt as optimiseur
 134         elif vt("1.13.0") <= vt(scipy.version.version) <= vt("1.13.99"):
 135             import daAlgorithms.Atoms.lbfgsb113hlt as optimiseur
 136         elif vt("1.14.0") <= vt(scipy.version.version) <= vt("1.14.99"):
 137             import daAlgorithms.Atoms.lbfgsb114hlt as optimiseur
 138         else:
 139             import scipy.optimize as optimiseur
 140         Minimum, J_optimal, Informations = optimiseur.fmin_l_bfgs_b(
 141             func        = CostFunction,
 142             x0          = Xini,
 143             fprime      = GradientOfCostFunction,
 144             args        = (),
 145             bounds      = selfA._parameters["Bounds"],
 146             maxfun      = selfA._parameters["MaximumNumberOfIterations"] - 1,
 147             factr       = selfA._parameters["CostDecrementTolerance"] * 1.e14,
 148             pgtol       = selfA._parameters["ProjectedGradientTolerance"],
 149             iprint      = selfA._parameters["optiprint"],
 150         )
 151         # nfeval = Informations['funcalls']
 152         # rc     = Informations['warnflag']
 153     elif selfA._parameters["Minimizer"] == "TNC":
 154         Minimum, nfeval, rc = scipy.optimize.fmin_tnc(
 155             func        = CostFunction,
 156             x0          = Xini,
 157             fprime      = GradientOfCostFunction,
 158             args        = (),
 159             bounds      = selfA._parameters["Bounds"],
 160             maxfun      = selfA._parameters["MaximumNumberOfIterations"],
 161             pgtol       = selfA._parameters["ProjectedGradientTolerance"],
 162             ftol        = selfA._parameters["CostDecrementTolerance"],
 163             messages    = selfA._parameters["optmessages"],
 164         )
 165     elif selfA._parameters["Minimizer"] == "CG":
 166         Minimum, fopt, nfeval, grad_calls, rc = scipy.optimize.fmin_cg(
 167             f           = CostFunction,
 168             x0          = Xini,
 169             fprime      = GradientOfCostFunction,
 170             args        = (),
 171             maxiter     = selfA._parameters["MaximumNumberOfIterations"],
 172             gtol        = selfA._parameters["GradientNormTolerance"],
 173             disp        = selfA._parameters["optdisp"],
 174             full_output = True,
 175         )
 176     elif selfA._parameters["Minimizer"] == "NCG":
 177         Minimum, fopt, nfeval, grad_calls, hcalls, rc = scipy.optimize.fmin_ncg(
 178             f           = CostFunction,
 179             x0          = Xini,
 180             fprime      = GradientOfCostFunction,
 181             args        = (),
 182             maxiter     = selfA._parameters["MaximumNumberOfIterations"],
 183             avextol     = selfA._parameters["CostDecrementTolerance"],
 184             disp        = selfA._parameters["optdisp"],
 185             full_output = True,
 186         )
 187     elif selfA._parameters["Minimizer"] == "BFGS":
 188         Minimum, fopt, gopt, Hopt, nfeval, grad_calls, rc = scipy.optimize.fmin_bfgs(
 189             f           = CostFunction,
 190             x0          = Xini,
 191             fprime      = GradientOfCostFunction,
 192             args        = (),
 193             maxiter     = selfA._parameters["MaximumNumberOfIterations"],
 194             gtol        = selfA._parameters["GradientNormTolerance"],
 195             disp        = selfA._parameters["optdisp"],
 196             full_output = True,
 197         )
 198     else:
 199         raise ValueError("Error in minimizer name: %s is unkown"%selfA._parameters["Minimizer"])
 200     #
 201     IndexMin = numpy.argmin( selfA.StoredVariables["CostFunctionJ"][nbPreviousSteps:] ) + nbPreviousSteps
 202     MinJ     = selfA.StoredVariables["CostFunctionJ"][IndexMin]
 203     #
 204     # Correction pour pallier a un bug de TNC sur le retour du Minimum
 205     # ----------------------------------------------------------------
 206     if selfA._parameters["StoreInternalVariables"] or selfA._toStore("CurrentState"):
 207         Minimum = selfA.StoredVariables["CurrentState"][IndexMin]
 208     #
 209     Xa = Minimum
 210     if __storeState:
 211         selfA._setInternalState("Xn", Xa)
 212     # --------------------------
 213     #
 214     selfA.StoredVariables["Analysis"].store( Xa )
 215     #
 216     if selfA._toStore("OMA") or \
 217             selfA._toStore("InnovationAtCurrentAnalysis") or \
 218             selfA._toStore("SigmaObs2") or \
 219             selfA._toStore("SimulationQuantiles") or \
 220             selfA._toStore("SimulatedObservationAtOptimum"):
 221         if selfA._toStore("SimulatedObservationAtCurrentState"):
 222             HXa = selfA.StoredVariables["SimulatedObservationAtCurrentState"][IndexMin]
 223         elif selfA._toStore("SimulatedObservationAtCurrentOptimum"):
 224             HXa = selfA.StoredVariables["SimulatedObservationAtCurrentOptimum"][-1]
 225         else:
 226             HXa = Hm( Xa )
 227         oma = Y - numpy.asarray(HXa).reshape((-1, 1))
 228     #
 229     if selfA._toStore("APosterioriCovariance") or \
 230             selfA._toStore("SimulationQuantiles") or \
 231             selfA._toStore("JacobianMatrixAtOptimum") or \
 232             selfA._toStore("KalmanGainAtOptimum"):
 233         HtM = HO["Tangent"].asMatrix(ValueForMethodForm = Xa)
 234         HtM = HtM.reshape(Y.size, Xa.size)  # ADAO & check shape
 235     if selfA._toStore("APosterioriCovariance") or \
 236             selfA._toStore("SimulationQuantiles") or \
 237             selfA._toStore("KalmanGainAtOptimum"):
 238         HaM = HO["Adjoint"].asMatrix(ValueForMethodForm = Xa)
 239         HaM = HaM.reshape(Xa.size, Y.size)  # ADAO & check shape
 240     if selfA._toStore("APosterioriCovariance") or \
 241             selfA._toStore("SimulationQuantiles"):
 242         A = HessienneEstimation(selfA, Xa.size, HaM, HtM, BI, RI)
 243     if selfA._toStore("APosterioriCovariance"):
 244         selfA.StoredVariables["APosterioriCovariance"].store( A )
 245     if selfA._toStore("JacobianMatrixAtOptimum"):
 246         selfA.StoredVariables["JacobianMatrixAtOptimum"].store( HtM )
 247     if selfA._toStore("KalmanGainAtOptimum"):
 248         if (Y.size <= Xb.size):
 249             KG  = B * HaM * (R + numpy.dot(HtM, B * HaM)).I
 250         elif (Y.size > Xb.size):
 251             KG = (BI + numpy.dot(HaM, RI * HtM)).I * HaM * RI
 252         selfA.StoredVariables["KalmanGainAtOptimum"].store( KG )
 253     #
 254     # Calculs et/ou stockages supplémentaires
 255     # ---------------------------------------
 256     if selfA._toStore("Innovation") or \
 257             selfA._toStore("SigmaObs2") or \
 258             selfA._toStore("MahalanobisConsistency") or \
 259             selfA._toStore("OMB"):
 260         Innovation  = Y - HXb
 261     if selfA._toStore("Innovation"):
 262         selfA.StoredVariables["Innovation"].store( Innovation )
 263     if selfA._toStore("BMA"):
 264         selfA.StoredVariables["BMA"].store( numpy.ravel(Xb) - numpy.ravel(Xa) )
 265     if selfA._toStore("OMA"):
 266         selfA.StoredVariables["OMA"].store( oma )
 267     if selfA._toStore("InnovationAtCurrentAnalysis"):
 268         selfA.StoredVariables["InnovationAtCurrentAnalysis"].store( oma )
 269     if selfA._toStore("OMB"):
 270         selfA.StoredVariables["OMB"].store( Innovation )
 271     if selfA._toStore("SigmaObs2"):
 272         TraceR = R.trace(Y.size)
 273         selfA.StoredVariables["SigmaObs2"].store( vfloat( (Innovation.T @ oma) ) / TraceR )
 274     if selfA._toStore("MahalanobisConsistency"):
 275         selfA.StoredVariables["MahalanobisConsistency"].store( float( 2. * MinJ / Innovation.size ) )
 276     if selfA._toStore("SimulationQuantiles"):
 277         QuantilesEstimations(selfA, A, Xa, HXa, Hm, HtM)
 278     if selfA._toStore("SimulatedObservationAtBackground"):
 279         selfA.StoredVariables["SimulatedObservationAtBackground"].store( HXb )
 280     if selfA._toStore("SimulatedObservationAtOptimum"):
 281         selfA.StoredVariables["SimulatedObservationAtOptimum"].store( HXa )
 282     #
 283     return 0
 284
 285 # ==============================================================================
 286 if __name__ == "__main__":
 287     print('\n AUTODIAGNOSTIC\n')