1
0
mirror of https://git.dev.opencascade.org/repos/occt.git synced 2025-04-10 18:51:21 +03:00
occt/src/math/math_NewtonMinimum.cxx
aml 94beb42a68 0029858: Modeling Data - Regression in GeomAPI_ExtremaCurveCurve
Fix Newton's minimum projection algorithm onto boundaries in case of conditional optimization.
Add possibility to detect several optimal points at initialization of the math_GlobOptMin.
2019-09-09 19:26:39 +03:00

271 lines
8.3 KiB
C++

// Created on: 1996-05-03
// Created by: Philippe MANGIN
// Copyright (c) 1996-1999 Matra Datavision
// Copyright (c) 1999-2014 OPEN CASCADE SAS
//
// This file is part of Open CASCADE Technology software library.
//
// This library is free software; you can redistribute it and/or modify it under
// the terms of the GNU Lesser General Public License version 2.1 as published
// by the Free Software Foundation, with special exception defined in the file
// OCCT_LGPL_EXCEPTION.txt. Consult the file LICENSE_LGPL_21.txt included in OCCT
// distribution for complete text of the license and disclaimer of any warranty.
//
// Alternatively, this file may be used under the terms of Open CASCADE
// commercial license or contractual agreement.
//#ifndef OCCT_DEBUG
#define No_Standard_RangeError
#define No_Standard_OutOfRange
#define No_Standard_DimensionError
//#endif
#include <math_Gauss.hxx>
#include <math_Jacobi.hxx>
#include <math_MultipleVarFunctionWithHessian.hxx>
#include <math_NewtonMinimum.hxx>
#include <Precision.hxx>
#include <Standard_DimensionError.hxx>
#include <StdFail_NotDone.hxx>
//=======================================================================
//function : math_NewtonMinimum
//purpose : Constructor
//=======================================================================
math_NewtonMinimum::math_NewtonMinimum(
const math_MultipleVarFunctionWithHessian& theFunction,
const Standard_Real theTolerance,
const Standard_Integer theNbIterations,
const Standard_Real theConvexity,
const Standard_Boolean theWithSingularity
)
: TheStatus (math_NotBracketed),
TheLocation(1, theFunction.NbVariables()),
TheGradient(1, theFunction.NbVariables()),
TheStep (1, theFunction.NbVariables(), 10.0 * theTolerance),
TheHessian (1, theFunction.NbVariables(), 1, theFunction.NbVariables()),
PreviousMinimum (0.0),
TheMinimum (0.0),
MinEigenValue (0.0),
XTol (theTolerance),
CTol (theConvexity),
nbiter (0),
NoConvexTreatement(theWithSingularity),
Convex (Standard_True),
myIsBoundsDefined(Standard_False),
myLeft(1, theFunction.NbVariables(), 0.0),
myRight(1, theFunction.NbVariables(), 0.0),
Done (Standard_False),
Itermax (theNbIterations)
{
}
//=======================================================================
//function : ~math_NewtonMinimum
//purpose : Destructor
//=======================================================================
math_NewtonMinimum::~math_NewtonMinimum()
{
}
//=======================================================================
//function : SetBoundary
//purpose : Set boundaries for conditional optimization
//=======================================================================
void math_NewtonMinimum::SetBoundary(const math_Vector& theLeftBorder,
const math_Vector& theRightBorder)
{
myLeft = theLeftBorder;
myRight = theRightBorder;
myIsBoundsDefined = Standard_True;
}
//=======================================================================
//function : Perform
//purpose :
//=======================================================================
void math_NewtonMinimum::Perform(math_MultipleVarFunctionWithHessian& F,
const math_Vector& StartingPoint)
{
math_Vector Point1 (1, F.NbVariables());
Point1 = StartingPoint;
math_Vector Point2(1, F.NbVariables());
math_Vector* precedent = &Point1;
math_Vector* suivant = &Point2;
math_Vector* auxiliaire = precedent;
Standard_Boolean Ok = Standard_True;
Standard_Integer NbConv = 0, ii, Nreduction;
Standard_Real VPrecedent, VItere;
Done = Standard_True;
TheStatus = math_OK;
nbiter = 0;
while ( Ok && (NbConv < 2) ) {
nbiter++;
// Positionnement
Ok = F.Values(*precedent, VPrecedent, TheGradient, TheHessian);
if (!Ok) {
Done = Standard_False;
TheStatus = math_FunctionError;
return;
}
if (nbiter==1) {
PreviousMinimum = VPrecedent;
TheMinimum = VPrecedent;
}
// Traitement de la non convexite
math_Jacobi CalculVP(TheHessian);
if ( !CalculVP.IsDone() ) {
Done = Standard_False;
TheStatus = math_FunctionError;
return;
}
MinEigenValue = CalculVP.Values() ( CalculVP.Values().Min());
if ( MinEigenValue < CTol)
{
Convex = Standard_False;
if (NoConvexTreatement && // Treatment is allowed.
Abs (MinEigenValue) > CTol) // Treatment will have effect.
{
Standard_Real Delta = CTol + 0.1 * Abs(MinEigenValue) - MinEigenValue;
for (ii=1; ii<=TheGradient.Length(); ii++)
TheHessian(ii, ii) += Delta;
}
else
{
Done = Standard_False;
TheStatus = math_FunctionError;
return;
}
}
// Schemas de Newton
math_Gauss LU(TheHessian, CTol/100);
if ( !LU.IsDone()) {
Done = Standard_False;
TheStatus = math_DirectionSearchError;
return;
}
LU.Solve(TheGradient, TheStep);
if (myIsBoundsDefined)
{
// Project point on bounds or nullify TheStep coords if point lies on boudary.
*suivant = *precedent - TheStep;
Standard_Real aMult = RealLast();
for(Standard_Integer anIdx = 1; anIdx <= myLeft.Upper(); anIdx++)
{
const Standard_Real anAbsStep = Abs(TheStep(anIdx));
if (anAbsStep < gp::Resolution())
continue;
if (suivant->Value(anIdx) < myLeft(anIdx))
{
Standard_Real aValue = Abs(precedent->Value(anIdx) - myLeft(anIdx)) / anAbsStep;
aMult = Min (aValue, aMult);
}
if (suivant->Value(anIdx) > myRight(anIdx))
{
Standard_Real aValue = Abs(precedent->Value(anIdx) - myRight(anIdx)) / anAbsStep;
aMult = Min (aValue, aMult);
}
}
if (aMult != RealLast())
{
if (aMult > Precision::PConfusion())
{
// Project point into param space.
TheStep *= aMult;
}
else
{
// Old point on border and new point out of border:
// Nullify corresponding TheStep indexes.
for(Standard_Integer anIdx = 1; anIdx <= myLeft.Upper(); anIdx++)
{
if ((Abs(precedent->Value(anIdx) - myRight(anIdx)) < Precision::PConfusion() && TheStep(anIdx) < 0.0) ||
(Abs(precedent->Value(anIdx) - myLeft(anIdx) ) < Precision::PConfusion() && TheStep(anIdx) > 0.0) )
{
TheStep(anIdx) = 0.0;
}
}
}
}
}
Standard_Boolean hasProblem = Standard_False;
do
{
*suivant = *precedent - TheStep;
// Gestion de la convergence
hasProblem = !(F.Value(*suivant, TheMinimum));
if (hasProblem)
{
TheStep /= 2.0;
}
} while (hasProblem);
if (IsConverged()) { NbConv++; }
else { NbConv=0; }
// Controle et corrections.
VItere = TheMinimum;
TheMinimum = PreviousMinimum;
Nreduction =0;
while (VItere > VPrecedent && Nreduction < 10) {
TheStep *= 0.4;
*suivant = *precedent - TheStep;
F.Value(*suivant, VItere);
Nreduction++;
}
if (VItere <= VPrecedent) {
auxiliaire = precedent;
precedent = suivant;
suivant = auxiliaire;
PreviousMinimum = VPrecedent;
TheMinimum = VItere;
Ok = (nbiter < Itermax);
if (!Ok && NbConv < 2) TheStatus = math_TooManyIterations;
}
else {
Ok = Standard_False;
TheStatus = math_DirectionSearchError;
}
}
TheLocation = *precedent;
}
//=======================================================================
//function : Dump
//purpose :
//=======================================================================
void math_NewtonMinimum::Dump(Standard_OStream& o) const
{
o<< "math_Newton Optimisation: ";
o << " Done =" << Done << std::endl;
o << " Status = " << (Standard_Integer)TheStatus << std::endl;
o << " Location Vector = " << Location() << std::endl;
o << " Minimum value = "<< Minimum()<< std::endl;
o << " Previous value = "<< PreviousMinimum << std::endl;
o << " Number of iterations = " <<NbIterations() << std::endl;
o << " Convexity = " << Convex << std::endl;
o << " Eigen Value = " << MinEigenValue << std::endl;
}