CGAL users discussion list

[DaniCamps <>]

Dear all,

I am finding that CGAL returns INFEASIBLE for a problem that Matlab
quadprog() can solve. I do not know if this is normal, or I am making a
mistake somewhere. I leave here the input of the program and would
appreciate if anyone can give me a hint.

BR

Daniel

Input of the program
=============

D
==
D[0][0]=0.000050
D[0][1]=0.000050
D[0][2]=0.000000
D[0][3]=0.000000
D[1][0]=0.000050
D[1][1]=0.000050
D[1][2]=0.000000
D[1][3]=0.000000
D[2][0]=0.000000
D[2][1]=0.000000
D[2][2]=0.000050
D[2][3]=0.000050
D[3][0]=0.000000
D[3][1]=0.000000
D[3][2]=0.000050
D[3][3]=0.000050

c:
==
c[0][0]=-0.002359
c[1][0]=-0.002359
c[2][0]=-0.002399
c[3][0]=-0.002399

A:
==
A[0][0]=1.000000
A[0][1]=0.000000
A[0][2]=1.000000
A[0][3]=0.000000
A[1][0]=0.000000
A[1][1]=1.000000
A[1][2]=0.000000
A[1][3]=1.000000

b:
==
b[0][0]=1.000000
b[1][0]=1.000000

ub:
==
ub[0][0]=0.914808
ub[1][0]=0.933087
ub[2][0]=1.061726
ub[3][0]=1.028364

lb:
==
lb[0][0]=0.000000
lb[1][0]=0.000000
lb[2][0]=0.000000
lb[3][0]=0.000000

The problem has equality constraints. Matlab quadprog() returns the
following solution:

Cost function
========
fval:[0][0]=-0.004717

x:
=
x[0][0]=0.293237
x[1][0]=0.306496
x[2][0]=0.706763
x[3][0]=0.693504

I configure the program in CGAL as follows:

// choose exact integral type
#ifdef CGAL_USE_GMP
#include <CGAL/Gmpzf.h>
typedef CGAL::Gmpzf ET;
#else
#include <CGAL/MP_Float.h>
typedef CGAL::MP_Float ET;
#endif

// program and solution types
typedef CGAL::Quadratic_program<double> Program;
typedef CGAL::Quadratic_program_solution<ET> Solution;

Solution s = CGAL::solve_quadratic_program(qp, ET());







--
Sent from: http://cgal-discuss.949826.n4.nabble.com/