SDP infeasibility and convergence issues

[Chen Chen]

Hello,

I have tried numerous solvers, but it seems my problem poses difficulty, likely because it's hard to find a strictly interior solution.  Unfortunately this results in claims of infeasibility when in fact I have provably feasible starting points (however, they are not strictly interior). Do you have any suggestions to alleviate this?

Thank you,

-Chen

[Johan Löfberg]

Typically hard to do anything about.

Try to add a margin to the constraint, and add a penalty on this margin in the objective, i.e., instead of G>=0 use G>=t*eye(n) and then penalize t with t^2, abs(t) etc

You are welcome to post (or mail) the model for further analysis

[Erling D. Andersen]

Have you tried MOSEK? You very welcome to email some examples. You can see how dump the examples to us at

http://users.isy.liu.se/johanl/yalmip/pmwiki.php?n=Solvers.MOSEK