1 # -*- coding: utf-8 -*-
3 # Copyright (C) 2008-2022 EDF R&D
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.
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.
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
19 # See http://www.salome-platform.org/ or email : webmaster.salome@opencascade.com
21 # Author: Jean-Philippe Argaud, jean-philippe.argaud@edf.fr, EDF R&D
24 3DVAR variational analysis with no inversion of B
26 __author__ = "Jean-Philippe ARGAUD"
28 import numpy, scipy, scipy.optimize, scipy.version
29 from daCore.NumericObjects import HessienneEstimation, QuantilesEstimations
30 from daCore.NumericObjects import RecentredBounds
31 from daCore.PlatformInfo import PlatformInfo
32 mpr = PlatformInfo().MachinePrecision()
34 # ==============================================================================
35 def van3dvar(selfA, Xb, Y, U, HO, CM, R, B, __storeState = False):
42 Hm = HO["Direct"].appliedTo
43 Ha = HO["Adjoint"].appliedInXTo
45 if HO["AppliedInX"] is not None and "HXb" in HO["AppliedInX"]:
46 HXb = numpy.asarray(Hm( Xb, HO["AppliedInX"]["HXb"] ))
48 HXb = numpy.asarray(Hm( Xb ))
49 HXb = HXb.reshape((-1,1))
50 if Y.size != HXb.size:
51 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))
52 if max(Y.shape) != max(HXb.shape):
53 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))
55 if selfA._toStore("JacobianMatrixAtBackground"):
56 HtMb = HO["Tangent"].asMatrix(ValueForMethodForm = Xb)
57 HtMb = HtMb.reshape(Y.size,Xb.size) # ADAO & check shape
58 selfA.StoredVariables["JacobianMatrixAtBackground"].store( HtMb )
62 if ("Bounds" in selfA._parameters) and selfA._parameters["Bounds"] is not None:
67 Xini = numpy.zeros(Xb.size)
69 # Définition de la fonction-coût
70 # ------------------------------
72 _V = numpy.asarray(v).reshape((-1,1))
73 _X = Xb + (B @ _V).reshape((-1,1))
74 if selfA._parameters["StoreInternalVariables"] or \
75 selfA._toStore("CurrentState") or \
76 selfA._toStore("CurrentOptimum"):
77 selfA.StoredVariables["CurrentState"].store( _X )
78 _HX = numpy.asarray(Hm( _X )).reshape((-1,1))
80 if selfA._toStore("SimulatedObservationAtCurrentState") or \
81 selfA._toStore("SimulatedObservationAtCurrentOptimum"):
82 selfA.StoredVariables["SimulatedObservationAtCurrentState"].store( _HX )
83 if selfA._toStore("InnovationAtCurrentState"):
84 selfA.StoredVariables["InnovationAtCurrentState"].store( _Innovation )
86 Jb = float( 0.5 * _V.T * (BT * _V) )
87 Jo = float( 0.5 * _Innovation.T * (RI * _Innovation) )
90 selfA.StoredVariables["CurrentIterationNumber"].store( len(selfA.StoredVariables["CostFunctionJ"]) )
91 selfA.StoredVariables["CostFunctionJb"].store( Jb )
92 selfA.StoredVariables["CostFunctionJo"].store( Jo )
93 selfA.StoredVariables["CostFunctionJ" ].store( J )
94 if selfA._toStore("IndexOfOptimum") or \
95 selfA._toStore("CurrentOptimum") or \
96 selfA._toStore("CostFunctionJAtCurrentOptimum") or \
97 selfA._toStore("CostFunctionJbAtCurrentOptimum") or \
98 selfA._toStore("CostFunctionJoAtCurrentOptimum") or \
99 selfA._toStore("SimulatedObservationAtCurrentOptimum"):
100 IndexMin = numpy.argmin( selfA.StoredVariables["CostFunctionJ"][nbPreviousSteps:] ) + nbPreviousSteps
101 if selfA._toStore("IndexOfOptimum"):
102 selfA.StoredVariables["IndexOfOptimum"].store( IndexMin )
103 if selfA._toStore("CurrentOptimum"):
104 selfA.StoredVariables["CurrentOptimum"].store( selfA.StoredVariables["CurrentState"][IndexMin] )
105 if selfA._toStore("SimulatedObservationAtCurrentOptimum"):
106 selfA.StoredVariables["SimulatedObservationAtCurrentOptimum"].store( selfA.StoredVariables["SimulatedObservationAtCurrentState"][IndexMin] )
107 if selfA._toStore("CostFunctionJbAtCurrentOptimum"):
108 selfA.StoredVariables["CostFunctionJbAtCurrentOptimum"].store( selfA.StoredVariables["CostFunctionJb"][IndexMin] )
109 if selfA._toStore("CostFunctionJoAtCurrentOptimum"):
110 selfA.StoredVariables["CostFunctionJoAtCurrentOptimum"].store( selfA.StoredVariables["CostFunctionJo"][IndexMin] )
111 if selfA._toStore("CostFunctionJAtCurrentOptimum"):
112 selfA.StoredVariables["CostFunctionJAtCurrentOptimum" ].store( selfA.StoredVariables["CostFunctionJ" ][IndexMin] )
115 def GradientOfCostFunction(v):
116 _V = numpy.asarray(v).reshape((-1,1))
117 _X = Xb + (B @ _V).reshape((-1,1))
118 _HX = numpy.asarray(Hm( _X )).reshape((-1,1))
120 GradJo = - BT * Ha( (_X, RI * (Y - _HX)) )
121 GradJ = numpy.ravel( GradJb ) + numpy.ravel( GradJo )
124 # Minimisation de la fonctionnelle
125 # --------------------------------
126 nbPreviousSteps = selfA.StoredVariables["CostFunctionJ"].stepnumber()
128 if selfA._parameters["Minimizer"] == "LBFGSB":
129 if "0.19" <= scipy.version.version <= "1.4.99":
130 import daAlgorithms.Atoms.lbfgsb14hlt as optimiseur
131 elif "1.5.0" <= scipy.version.version <= "1.7.99":
132 import daAlgorithms.Atoms.lbfgsb17hlt as optimiseur
133 elif "1.8.0" <= scipy.version.version <= "1.8.99":
134 import daAlgorithms.Atoms.lbfgsb18hlt as optimiseur
135 elif "1.9.0" <= scipy.version.version <= "1.9.99":
136 import daAlgorithms.Atoms.lbfgsb19hlt as optimiseur
138 import scipy.optimize as optimiseur
139 Minimum, J_optimal, Informations = optimiseur.fmin_l_bfgs_b(
142 fprime = GradientOfCostFunction,
144 bounds = RecentredBounds(selfA._parameters["Bounds"], Xb, BI),
145 maxfun = selfA._parameters["MaximumNumberOfIterations"]-1,
146 factr = selfA._parameters["CostDecrementTolerance"]*1.e14,
147 pgtol = selfA._parameters["ProjectedGradientTolerance"],
148 iprint = selfA._parameters["optiprint"],
150 # nfeval = Informations['funcalls']
151 # rc = Informations['warnflag']
152 elif selfA._parameters["Minimizer"] == "TNC":
153 Minimum, nfeval, rc = scipy.optimize.fmin_tnc(
156 fprime = GradientOfCostFunction,
158 bounds = RecentredBounds(selfA._parameters["Bounds"], Xb, BI),
159 maxfun = selfA._parameters["MaximumNumberOfIterations"],
160 pgtol = selfA._parameters["ProjectedGradientTolerance"],
161 ftol = selfA._parameters["CostDecrementTolerance"],
162 messages = selfA._parameters["optmessages"],
164 elif selfA._parameters["Minimizer"] == "CG":
165 Minimum, fopt, nfeval, grad_calls, rc = scipy.optimize.fmin_cg(
168 fprime = GradientOfCostFunction,
170 maxiter = selfA._parameters["MaximumNumberOfIterations"],
171 gtol = selfA._parameters["GradientNormTolerance"],
172 disp = selfA._parameters["optdisp"],
175 elif selfA._parameters["Minimizer"] == "NCG":
176 Minimum, fopt, nfeval, grad_calls, hcalls, rc = scipy.optimize.fmin_ncg(
179 fprime = GradientOfCostFunction,
181 maxiter = selfA._parameters["MaximumNumberOfIterations"],
182 avextol = selfA._parameters["CostDecrementTolerance"],
183 disp = selfA._parameters["optdisp"],
186 elif selfA._parameters["Minimizer"] == "BFGS":
187 Minimum, fopt, gopt, Hopt, nfeval, grad_calls, rc = scipy.optimize.fmin_bfgs(
190 fprime = GradientOfCostFunction,
192 maxiter = selfA._parameters["MaximumNumberOfIterations"],
193 gtol = selfA._parameters["GradientNormTolerance"],
194 disp = selfA._parameters["optdisp"],
198 raise ValueError("Error in minimizer name: %s is unkown"%selfA._parameters["Minimizer"])
200 IndexMin = numpy.argmin( selfA.StoredVariables["CostFunctionJ"][nbPreviousSteps:] ) + nbPreviousSteps
201 MinJ = selfA.StoredVariables["CostFunctionJ"][IndexMin]
203 # Correction pour pallier a un bug de TNC sur le retour du Minimum
204 # ----------------------------------------------------------------
205 if selfA._parameters["StoreInternalVariables"] or selfA._toStore("CurrentState"):
206 Minimum = selfA.StoredVariables["CurrentState"][IndexMin]
208 Minimum = Xb + B * Minimum.reshape((-1,1)) # Pas @
211 if __storeState: selfA._setInternalState("Xn", Xa)
212 #--------------------------
214 selfA.StoredVariables["Analysis"].store( Xa )
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]
227 oma = Y - HXa.reshape((-1,1))
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"):
243 A = HessienneEstimation(selfA, Xa.size, HaM, HtM, BI, RI)
244 if selfA._toStore("APosterioriCovariance"):
245 selfA.StoredVariables["APosterioriCovariance"].store( A )
246 if selfA._toStore("JacobianMatrixAtOptimum"):
247 selfA.StoredVariables["JacobianMatrixAtOptimum"].store( HtM )
248 if selfA._toStore("KalmanGainAtOptimum"):
249 if (Y.size <= Xb.size): KG = B * HaM * (R + numpy.dot(HtM, B * HaM)).I
250 elif (Y.size > Xb.size): KG = (BI + numpy.dot(HaM, RI * HtM)).I * HaM * RI
251 selfA.StoredVariables["KalmanGainAtOptimum"].store( KG )
253 # Calculs et/ou stockages supplémentaires
254 # ---------------------------------------
255 if selfA._toStore("Innovation") or \
256 selfA._toStore("SigmaObs2") or \
257 selfA._toStore("MahalanobisConsistency") or \
258 selfA._toStore("OMB"):
260 if selfA._toStore("Innovation"):
261 selfA.StoredVariables["Innovation"].store( Innovation )
262 if selfA._toStore("BMA"):
263 selfA.StoredVariables["BMA"].store( numpy.ravel(Xb) - numpy.ravel(Xa) )
264 if selfA._toStore("OMA"):
265 selfA.StoredVariables["OMA"].store( oma )
266 if selfA._toStore("InnovationAtCurrentAnalysis"):
267 selfA.StoredVariables["InnovationAtCurrentAnalysis"].store( oma )
268 if selfA._toStore("OMB"):
269 selfA.StoredVariables["OMB"].store( Innovation )
270 if selfA._toStore("SigmaObs2"):
271 TraceR = R.trace(Y.size)
272 selfA.StoredVariables["SigmaObs2"].store( float( (Innovation.T @ oma) ) / TraceR )
273 if selfA._toStore("MahalanobisConsistency"):
274 selfA.StoredVariables["MahalanobisConsistency"].store( float( 2.*MinJ/Innovation.size ) )
275 if selfA._toStore("SimulationQuantiles"):
276 QuantilesEstimations(selfA, A, Xa, HXa, Hm, HtM)
277 if selfA._toStore("SimulatedObservationAtBackground"):
278 selfA.StoredVariables["SimulatedObservationAtBackground"].store( HXb )
279 if selfA._toStore("SimulatedObservationAtOptimum"):
280 selfA.StoredVariables["SimulatedObservationAtOptimum"].store( HXa )
284 # ==============================================================================
285 if __name__ == "__main__":
286 print('\n AUTODIAGNOSTIC\n')