issue with "max" operator in constraint

[Olivier Mégel]

Hi,

I came across a strange behavior when using the "max" operator in a constraint. It looks like the operator is confused when one term is an sdpvar, and the other one a double:

options=sdpsettings('verbose',0);

a=sdpvar(1,1);

b=sdpvar(1,1);

C=[];

C=[C,a==1];

C=[C,b==max(-1*a,0)];

sol = optimize(C,[],options)

double(a)

double(b)

sol = 

    yalmiptime: 0.3557

    solvertime: 0.0013

          info: 'Successfully solved (GUROBI-GUROBI)'

       problem: 0

          time: 0.0013

ans =

     1

ans =

   -1.0000

(the value of b is -1, but it should be max(-1*1,0) , i.e 0). However, if I do the following modification, it works (down to the solver numerical precision):

options=sdpsettings('verbose',0);

a=sdpvar(1,1);

b=sdpvar(1,1);

DummyZero=sdpvar(1,1);

C=[];

C=[C,a==1];

C=[C,b==max(-1*a,DummyZero)];

C=[C,DummyZero==0];

sol = optimize(C,[],options)

double(a)

double(b)

sol = 

    yalmiptime: 0.3967

    solvertime: 0.0013

          info: 'Successfully solved (GUROBI-GUROBI)'

       problem: 0

          time: 0.0013

ans =

     1

ans =

  -2.5466e-11

Am I doing something wrong, or is this a bug ? (note that I can live with the DummyZero sdpvar if I have to).

Best,

Olivier

[Johan Löfberg]

That's a bug in the non-convex case of the max operator, will fix.

For now, just comment out line 44 in max_internal

[Johan Löfberg]

meant line 44-53