why yalmip can't recognize the problem is a MISOCP problem???

[马彦锴]

Dear Johan,

        I am using yalmip and GUROBI to solve a MISOCP? However, the matlab says "Model is infeasible or unbounded"

        The variable x is pure integer variable and variable t is a quadratic variable. I think it is a typical MISOCP. I will appreciate it if you help me solve the problem.

        

[Michal Adamaszek]

MOSEK gives you the same answer. I think the minus in the Objective should not be there, given that t is the root variable in the cone and you probably want to minimize t.

[Johan Löfberg]

it's unbounded  so you can create arbitrarily good solutions

which is pretty obvious as you re maximizing t were t>=something, hence t=inf is optimal

(as michal said...)

[马彦锴]

It did work when I minimized the objective. however my original plan is maximizing it indeed. maybe I am missing some constraints. Thank you very much.

[马彦锴]

Thank you very much. It is so kind of you helping me solve the problem so many times.

[Johan Löfberg]

Maximizing a convex quadrratic function is a nonconvex problem, so you cannot use some simple epigraph reformulation. If you want ot maximize q(x), you could write it as maximize t subject to q(x)>=t, but that constraint is a nonconvex constraint if q(x) is convex

[马彦锴]

Ok,I understand. Maximizing it is the boundary of the cone. Maybe I should reformulate it in another way. Thank you for all