cplexqp, when used for nonconvex quadratic programming, is giving the same results (fvalue) when tried with different random initial points. Actually, different initial points should make the algorithm converge to different local minima, I believe? Why doesnt this happen?

Thanks in advance

Topic