problem about warnings of gurobi and mosek in socp

Ziliang Lin

unread,

Nov 27, 2018, 10:29:56 PM11/27/18

to YALMIP

Hey guys,

i am trying to solve a socp problem by gurobi. I run my code successfully but there comes a warning that :"model.cones is deprecated and will be removed in a future version.Use model.quadcon instead. " does this warning have any effect in my result? what should i do to solve it?

Additionally, when i solve the same problem by mosek ,the result is "mosek error:MSK_RES_TRM_STALL () " and "Numerical problems" . what is that mean?

anybody help me please...

code.m

Johan Löfberg

unread,

Nov 28, 2018, 1:34:37 AM11/28/18

to YALMIP

gurobi, no that's just an heads-up for developers

mosek has terminated with numerical problems and a significant gap. You will have to proceed with caution and see if the solution really is relevant for you.

Primal. obj: -8.0566850999e+05 nrm: 7e+04 Viol. con: 4e-01 var: 3e-02 cones: 0e+00

Dual. obj: -8.0524369316e+05 nrm: 4e+05 Viol. con: 0e+00 var: 3e-01 cones: 0e+00

Your model appears to be badly scaled. For instance, you have coefficients in your objective spanning something like 6 orders of magnitudes

>> fun

Quadratic scalar (real, 144 variables)

Current value: 805242.8415

Coeffiecient range: 0.0004 to 1353.0816

and you have constraints with silly small coefficients, i.e. you are adding/comparing things of enormously different scale

+++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++

>> F(49)

++++++++++++++++++++++++++++++++++++++++++++++++++++++++

| ID| Constraint| Coefficient range|

++++++++++++++++++++++++++++++++++++++++++++++++++++++++

| #1| Equality constraint 24x1| 5.1901e-08 to 1|

++++++++++++++++++++++++++++++++++++++++++++++++++++++++

Ziliang Lin

unread,

Nov 28, 2018, 5:23:16 AM11/28/18

to YALMIP

All right. Thank you very much!

Reply all

Reply to author

Forward