Does anyone know how to solve MI-SDP problem in Julia?

I am working on an MI-SDP model and I have no idea how to solve it using Julia. I tried solving the model in convex using MOSEK as a solver, it turns out it failed, throwing some nasty error.

Any ideas on solving Mixed Integer SDP in julia is appreciated and thanks!!

1 Like

I can recomend the Pajarito solver. It combines a MIP solver and an SDP solver to solve MISDP.

1 Like

Thanks, really thanks. So I access pajarito through convex or JuMP?

I use JuMP, but I believe it connects to convex as well.

I am trying to solve MI-SDP in PowerModels with JuMP 19.2. After trying different packages, my observations are followings. correct me if I am wrong.

  • Pajarito and SCIP are yet to support this JuMP version
  • Juniper does not support constraints written in JuMP.secondordercone format
  • Mosek can not solve MISDP
    Is there a workaround or other packages I can try? Any help is appreciated. Thank you!

SCIP itself should work fine with JuMP 0.19.2; do you mean that SCIP-SDP has not been wrapped?

Yes, you are right. I made a mistake. SCIP supports the JuMP version but I am not sure if SCIP-SDP is available in Julia.

Do you know if there are any developments in this area?
It seems to me that Pajarito can still not be used.

No. Help updating Pajarito also appreciated.

Sorry for bothering you two different places. I would like to help. Have no experience developing other people’s code though. How can I help? Are there any guidelines for contributing, or explanation about the structure of your package?

Hello everyone,
Is there any update regarding this topic?
Appreciate any information.

Pajarito has been updated and supports the latest version of JuMP

1 Like

Thank you, @odow, for your response. Is there any preference for a conic solver and a MIP solver for Pajarito? I am using Mosek and SCIP now, but I am unsure if it is the best choice.

Mosek and SCIP are good choices. You could also try a commercial MIP solver like Gurobi or CPLEX.