1 #-*-coding:iso-8859-1-*-
3 # Copyright (C) 2008-2013 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 from daCore import BasicObjects, PlatformInfo
25 m = PlatformInfo.SystemUsage()
29 # ==============================================================================
30 class ElementaryAlgorithm(BasicObjects.Algorithm):
32 BasicObjects.Algorithm.__init__(self, "GRADIENTTEST")
33 self.defineRequiredParameter(
34 name = "ResiduFormula",
37 message = "Formule de résidu utilisée",
38 listval = ["Norm", "Taylor"],
40 self.defineRequiredParameter(
41 name = "EpsilonMinimumExponent",
44 message = "Exposant minimal en puissance de 10 pour le multiplicateur d'incrément",
48 self.defineRequiredParameter(
49 name = "InitialDirection",
52 message = "Direction initiale de la dérivée directionnelle autour du point nominal",
54 self.defineRequiredParameter(
55 name = "AmplitudeOfInitialDirection",
58 message = "Amplitude de la direction initiale de la dérivée directionnelle autour du point nominal",
60 self.defineRequiredParameter(
62 typecast = numpy.random.seed,
63 message = "Graine fixée pour le générateur aléatoire",
65 self.defineRequiredParameter(
69 message = "Trace et sauve les résultats",
71 self.defineRequiredParameter(
73 default = self._name+"_result_file",
75 message = "Nom de base (hors extension) des fichiers de sauvegarde des résultats",
77 self.defineRequiredParameter(
81 message = "Titre du tableau et de la figure",
83 self.defineRequiredParameter(
87 message = "Label de la courbe tracée dans la figure",
90 def run(self, Xb=None, Y=None, U=None, HO=None, EM=None, CM=None, R=None, B=None, Q=None, Parameters=None):
91 logging.debug("%s Lancement"%self._name)
92 logging.debug("%s Taille mémoire utilisée de %.1f Mo"%(self._name, m.getUsedMemory("M")))
94 # Paramètres de pilotage
95 # ----------------------
96 self.setParameters(Parameters)
98 # Opérateur d'observation
99 # -----------------------
100 Hm = HO["Direct"].appliedTo
101 if self._parameters["ResiduFormula"] is "Taylor":
102 Ht = HO["Tangent"].appliedInXTo
104 # Construction des perturbations
105 # ------------------------------
106 Perturbations = [ 10**i for i in xrange(self._parameters["EpsilonMinimumExponent"],1) ]
107 Perturbations.reverse()
109 # Calcul du point courant
110 # -----------------------
111 X = numpy.asmatrix(numpy.ravel( Xb )).T
112 FX = numpy.asmatrix(numpy.ravel( Hm( X ) )).T
113 FX = numpy.asmatrix(numpy.ravel( FX )).T
114 NormeX = numpy.linalg.norm( X )
115 NormeFX = numpy.linalg.norm( FX )
117 # Fabrication de la direction de l'incrément dX
118 # ----------------------------------------------
119 if len(self._parameters["InitialDirection"]) == 0:
123 dX0.append( numpy.random.normal(0.,abs(v)) )
125 dX0.append( numpy.random.normal(0.,X.mean()) )
127 dX0 = numpy.ravel( self._parameters["InitialDirection"] )
129 dX0 = float(self._parameters["AmplitudeOfInitialDirection"]) * numpy.matrix( dX0 ).T
131 # Calcul du gradient au point courant X pour l'incrément dX
132 # ---------------------------------------------------------
133 if self._parameters["ResiduFormula"] is "Taylor":
134 GradFxdX = Ht( (X, dX0) )
135 GradFxdX = numpy.asmatrix(numpy.ravel( GradFxdX )).T
137 # Entete des resultats
138 # --------------------
139 if self._parameters["ResiduFormula"] is "Taylor":
141 On observe le residu issu du développement de Taylor de la fonction F :
143 R(Alpha) = || F(X+Alpha*dX) - F(X) - Alpha * TangentF_X(dX) ||
145 Ce résidu doit décroître en Alpha**2 selon Alpha.
146 On prend dX0 = Normal(0,X) et dX = Alpha*dX0. F est le code de calcul.
148 elif self._parameters["ResiduFormula"] is "Norm":
150 On observe le residu, qui est une approximation du gradient :
152 || F(X+Alpha*dX) - F(X) ||
153 R(Alpha) = ---------------------------
156 qui doit rester constant jusqu'à ce qu'on atteigne la précision du calcul.
157 On prend dX0 = Normal(0,X) et dX = Alpha*dX0. F est le code de calcul.
162 if len(self._parameters["ResultTitle"]) > 0:
163 msgs = " ====" + "="*len(self._parameters["ResultTitle"]) + "====\n"
164 msgs += " " + self._parameters["ResultTitle"] + "\n"
165 msgs += " ====" + "="*len(self._parameters["ResultTitle"]) + "====\n"
170 msg = " i Alpha ||X|| ||F(X)|| ||F(X+dX)|| ||dX|| ||F(X+dX)-F(X)|| ||F(X+dX)-F(X)||/||dX|| R(Alpha) "
172 msgs += "\n" + "-"*nbtirets
174 msgs += "\n" + "-"*nbtirets
184 # Boucle sur les perturbations
185 # ----------------------------
186 for i,amplitude in enumerate(Perturbations):
189 FX_plus_dX = Hm( X + dX )
190 FX_plus_dX = numpy.asmatrix(numpy.ravel( FX_plus_dX )).T
192 NormedX = numpy.linalg.norm( dX )
193 NormeFXdX = numpy.linalg.norm( FX_plus_dX )
194 NormedFX = numpy.linalg.norm( FX_plus_dX - FX )
195 NormedFXsdX = NormedFX/NormedX
197 if self._parameters["ResiduFormula"] is "Taylor":
198 NormedFXGdX = numpy.linalg.norm( FX_plus_dX - FX - amplitude * GradFxdX )
200 NormedFXsAm = NormedFX/amplitude
202 # if numpy.abs(NormedFX) < 1.e-20:
205 NormesdX.append( NormedX )
206 NormesFXdX.append( NormeFXdX )
207 NormesdFX.append( NormedFX )
208 if self._parameters["ResiduFormula"] is "Taylor":
209 NormesdFXGdX.append( NormedFXGdX )
210 NormesdFXsdX.append( NormedFXsdX )
211 NormesdFXsAm.append( NormedFXsAm )
213 if self._parameters["ResiduFormula"] is "Taylor":
215 elif self._parameters["ResiduFormula"] is "Norm":
217 if Normalisation < 0 : Normalisation = Residu
219 msg = " %2i %5.0e %9.3e %9.3e %9.3e %9.3e %9.3e | %9.3e | %9.3e"%(i,amplitude,NormeX,NormeFX,NormeFXdX,NormedX,NormedFX,NormedFXsdX,Residu)
222 self.StoredVariables["CostFunctionJ"].store( Residu )
223 msgs += "\n" + "-"*nbtirets
226 # Sorties eventuelles
227 # -------------------
229 print "Results of gradient stability check:"
232 if self._parameters["PlotAndSave"]:
233 f = open(str(self._parameters["ResultFile"])+".txt",'a')
237 Residus = self.StoredVariables["CostFunctionJ"][-len(Perturbations):]
238 if self._parameters["ResiduFormula"] is "Taylor":
239 PerturbationsCarre = [ 10**(2*i) for i in xrange(-len(NormesdFXGdX)+1,1) ]
240 PerturbationsCarre.reverse()
244 titre = self._parameters["ResultTitle"],
245 label = self._parameters["ResultLabel"],
248 filename = str(self._parameters["ResultFile"])+".ps",
249 YRef = PerturbationsCarre,
250 normdY0 = numpy.log10( NormesdFX[0] ),
252 elif self._parameters["ResiduFormula"] is "Norm":
256 titre = self._parameters["ResultTitle"],
257 label = self._parameters["ResultLabel"],
260 filename = str(self._parameters["ResultFile"])+".ps",
263 logging.debug("%s Taille mémoire utilisée de %.1f Mo"%(self._name, m.getUsedMemory("M")))
264 logging.debug("%s Terminé"%self._name)
268 # ==============================================================================
279 YRef = None, # Vecteur de reference a comparer a Y
280 recalYRef = True, # Decalage du point 0 de YRef à Y[0]
281 normdY0 = 0., # Norme de DeltaY[0]
285 __g = __gnuplot.Gnuplot(persist=1) # persist=1
286 # __g('set terminal '+__gnuplot.GnuplotOpts.default_term)
287 __g('set style data lines')
290 __g('set title "'+titre+'"')
291 # __g('set xrange [] reverse')
292 # __g('set yrange [0:2]')
295 steps = numpy.log10( X )
296 __g('set xlabel "Facteur multiplicatif de dX, en echelle log10"')
299 __g('set xlabel "Facteur multiplicatif de dX"')
302 values = numpy.log10( Y )
303 __g('set ylabel "Amplitude du residu, en echelle log10"')
306 __g('set ylabel "Amplitude du residu"')
308 __g.plot( __gnuplot.Data( steps, values, title=label, with_='lines lw 3' ) )
311 valuesRef = numpy.log10( YRef )
314 if recalYRef and not numpy.all(values < -8):
315 valuesRef = valuesRef + values[0]
316 elif recalYRef and numpy.all(values < -8):
317 valuesRef = valuesRef + normdY0
320 __g.replot( __gnuplot.Data( steps, valuesRef, title="Reference", with_='lines lw 1' ) )
322 if filename is not "":
323 __g.hardcopy( filename, color=1)
325 raw_input('Please press return to continue...\n')
327 # ==============================================================================
328 if __name__ == "__main__":
329 print '\n AUTODIAGNOSTIC \n'