mirror of
https://git.dev.opencascade.org/repos/occt.git
synced 2025-08-09 13:22:24 +03:00
0025635: Wrong result of 2D-extrema between two ellipsis
Fixed Lipschitz constant evaluation in case co-parametrized objects. Fixed 2dextrema output. Testcase update to new behavior. Test cases for issue CR25635 Correction of test cases for issue CR25635
This commit is contained in:
@@ -23,6 +23,7 @@
|
||||
#include <math_Powell.hxx>
|
||||
#include <Standard_Integer.hxx>
|
||||
#include <Standard_Real.hxx>
|
||||
#include <Precision.hxx>
|
||||
|
||||
|
||||
//=======================================================================
|
||||
@@ -194,9 +195,6 @@ void math_GlobOptMin::Perform()
|
||||
{
|
||||
Standard_Integer i;
|
||||
|
||||
// Compute initial values for myF, myY, myC.
|
||||
computeInitialValues();
|
||||
|
||||
// Compute parameters range
|
||||
Standard_Real minLength = RealLast();
|
||||
Standard_Real maxLength = RealFirst();
|
||||
@@ -209,6 +207,18 @@ void math_GlobOptMin::Perform()
|
||||
maxLength = currentLength;
|
||||
}
|
||||
|
||||
if (minLength < Precision::PConfusion())
|
||||
{
|
||||
#ifdef OCCT_DEBUG
|
||||
cout << "math_GlobOptMin::Perform(): Degenerated parameters space" << endl;
|
||||
#endif
|
||||
|
||||
return;
|
||||
}
|
||||
|
||||
// Compute initial values for myF, myY, myC.
|
||||
computeInitialValues();
|
||||
|
||||
myE1 = minLength * myTol;
|
||||
myE2 = maxLength * myTol;
|
||||
if (myC > 1.0)
|
||||
@@ -293,7 +303,7 @@ void math_GlobOptMin::computeInitialValues()
|
||||
Standard_Integer i;
|
||||
math_Vector aCurrPnt(1, myN);
|
||||
math_Vector aBestPnt(1, myN);
|
||||
|
||||
math_Vector aParamStep(1, myN);
|
||||
Standard_Real aCurrVal = RealLast();
|
||||
Standard_Real aBestVal = RealLast();
|
||||
|
||||
@@ -324,21 +334,29 @@ void math_GlobOptMin::computeInitialValues()
|
||||
mySolCount++;
|
||||
|
||||
// Lipschitz const approximation
|
||||
Standard_Real aLipConst = 0.0, aPrevVal;
|
||||
Standard_Real aLipConst = 0.0, aPrevValDiag, aPrevValProj;
|
||||
Standard_Integer aPntNb = 13;
|
||||
myFunc->Value(myA, aPrevVal);
|
||||
myFunc->Value(myA, aPrevValDiag);
|
||||
aPrevValProj = aPrevValDiag;
|
||||
Standard_Real aStep = (myB - myA).Norm() / aPntNb;
|
||||
aParamStep = (myB - myA) / aPntNb;
|
||||
for(i = 1; i <= aPntNb; i++)
|
||||
{
|
||||
aCurrPnt = myA + (myB - myA) * i / (aPntNb - 1);
|
||||
aCurrPnt = myA + aParamStep * i;
|
||||
|
||||
// Walk over diagonal.
|
||||
myFunc->Value(aCurrPnt, aCurrVal);
|
||||
aLipConst = Max (Abs(aCurrVal - aPrevValDiag), aLipConst);
|
||||
aPrevValDiag = aCurrVal;
|
||||
|
||||
if(Abs(aCurrVal - aPrevVal) / aStep > aLipConst)
|
||||
aLipConst = Abs(aCurrVal - aPrevVal) / aStep;
|
||||
|
||||
aPrevVal = aCurrVal;
|
||||
// Walk over diag in projected space aPnt(1) = myA(1) = const.
|
||||
aCurrPnt(1) = myA(1);
|
||||
myFunc->Value(aCurrPnt, aCurrVal);
|
||||
aLipConst = Max (Abs(aCurrVal - aPrevValProj), aLipConst);
|
||||
aPrevValProj = aCurrVal;
|
||||
}
|
||||
aLipConst *= Sqrt(myN);
|
||||
|
||||
aLipConst *= Sqrt(myN) / aStep;
|
||||
|
||||
if (aLipConst < myC * 0.1)
|
||||
{
|
||||
|
Reference in New Issue
Block a user