Problem with QP example from quadprog.

Tue May 19, 2015 9:03 pm

So I ran the code http://www.labri.fr/perso/guenneba/code/QuadProg/ (changing internal:: by std::) and it all works as expected: I get

Now, I try to be a bit bolder and run the example here http://www.inside-r.org/packages/cran/quadprog/docs/solve.QP, which is:

min! -(0 5 0) %*% b + 1/2 b^T b
u.c.: A^T b >= b0
with b0 = (-8,2,0)^T
and
....(-4 2 0)
A=(-3 1 -2)
....( 0 0 1)

(a better edited formulation of the toy problem is at the link above)

(removing the copyrights to save space), I think, this gives the following code:

#include <iostream>
#include <Eigen/Dense>
#include "eiquadprog.hpp"

using namespace Eigen;

template<typename Vec, typename Mat> void foo() {
Mat G(3,3);
Vec g0(3);
Mat CE(3,1);
Vec ce0(1);
Mat CI(3,3);
Vec ci0(3);
Vec x(3);

G(0,0)=1.0; G(0,1)=0.0; G(0,2)=0.0;
G(1,0)=0.0; G(1,1)=1.0; G(1,2)=0.0;
G(2,0)=0.0; G(2,1)=0.0; G(2,2)=1.0;

g0(0)=0.0;g0(1)=-5.0;g0(2)=0.0;

CE(0,0)=0.0;
CE(1,0)=0.0;
CE(2,0)=0.0;

ce0(0)=0.0;

CI(0,0)=-4.0;CI(0,1)=2.0;CI(0,2)=0.0;
CI(1,0)=-3.0;CI(1,1)=1.0;CI(1,2)=-2.0;
CI(2,0)=0.0;CI(2,1)=0.0;CI(2,2)=1.0;

ci0(0)=8.0;ci0(1)=-2.0;ci0(2)=0.0;

std::cout << "f: " << solve_quadprog(G, g0, CE, ce0, CI, ci0, x) << std::endl;
std::cout << "x: ";
for (int i = 0; i < x.size(); i++)
std::cout << x(i) << ' ';
std::cout << std::endl;
}

int main(int argc, char** argv){

foo<Eigen::VectorXd,MatrixXd>();

}

But when I run this I get

where the correct solution should be:

I tried transposing the CI matrix in the eigen-example code
but could still not get the same solutions as those from the quadprog example.
Probably I'm setting the problem wrongly and I was wondering whether someone
has encountered a similar problem setting up a quadprog code in
eigquadprog and could point to what I am doing wrong.

vvakk Registered Member Posts 2 Karma 0	Problem with QP example from quadprog. Tue May 19, 2015 9:03 pm So I ran the code http://www.labri.fr/perso/guenneba/code/QuadProg/ (changing internal:: by std::) and it all works as expected: I get Code: Select all `f: 6.4 x: 4.23793e-17 2 -7.67063e-17` Now, I try to be a bit bolder and run the example here http://www.inside-r.org/packages/cran/quadprog/docs/solve.QP, which is: min! -(0 5 0) %% b + 1/2 b^T b u.c.: A^T b >= b0 with b0 = (-8,2,0)^T and ....(-4 2 0) A=(-3 1 -2) ....( 0 0 1) (a better edited formulation of the toy problem is at the link above) (removing the copyrights to save space), I think, this gives the following code: #include <iostream> #include <Eigen/Dense> #include "eiquadprog.hpp" using namespace Eigen; template<typename Vec, typename Mat> void foo() { Mat G(3,3); Vec g0(3); Mat CE(3,1); Vec ce0(1); Mat CI(3,3); Vec ci0(3); Vec x(3); G(0,0)=1.0; G(0,1)=0.0; G(0,2)=0.0; G(1,0)=0.0; G(1,1)=1.0; G(1,2)=0.0; G(2,0)=0.0; G(2,1)=0.0; G(2,2)=1.0; g0(0)=0.0;g0(1)=-5.0;g0(2)=0.0; CE(0,0)=0.0; CE(1,0)=0.0; CE(2,0)=0.0; ce0(0)=0.0; CI(0,0)=-4.0;CI(0,1)=2.0;CI(0,2)=0.0; CI(1,0)=-3.0;CI(1,1)=1.0;CI(1,2)=-2.0; CI(2,0)=0.0;CI(2,1)=0.0;CI(2,2)=1.0; ci0(0)=8.0;ci0(1)=-2.0;ci0(2)=0.0; std::cout << "f: " << solve_quadprog(G, g0, CE, ce0, CI, ci0, x) << std::endl; std::cout << "x: "; for (int i = 0; i < x.size(); i++) std::cout << x(i) << ' '; std::cout << std::endl; } int main(int argc, char* argv){ foo<Eigen::VectorXd,MatrixXd>(); } But when I run this I get Code: Select all `f: -12.5 x: 0 5 0` where the correct solution should be: Code: Select all `f: -2.380952 x: 0.4761905 1.0476190 2.0952381` I tried transposing the CI matrix in the eigen-example code but could still not get the same solutions as those from the quadprog example. Probably I'm setting the problem wrongly and I was wondering whether someone has encountered a similar problem setting up a quadprog code in eigquadprog and could point to what I am doing wrong.
ggael Moderator Posts 3447 Karma 19 OS	Re: Problem with QP example from quadprog. Mon May 25, 2015 9:16 pm CE and ce0 should be null: CE.resize(3,0); ce0.resize(0); then set the other entries as follows: Code: Select all `G.setIdentity(); g0 << 0, -5, 0; CI.resize(3,3); CI << -4, 2, 0, -3, 1, -2, 0, 0, 1; ci0.resize(3); ci0 << 8, -2, 0; solve_quadprog(G, g0, CE, ce0, CI, ci0, x)` btw, I've updated the eiquadprog.hpp file to work with eigen 3.2.
vvakk Registered Member Posts 2 Karma 0	Re: Problem with QP example from quadprog. Tue May 26, 2015 2:34 am Thanks a lot!

Problem with QP example from quadprog.

Page 1 of 1 (3 posts)

Problem with QP example from quadprog.

Re: Problem with QP example from quadprog.

Re: Problem with QP example from quadprog.

Bookmarks

Who is online