LMILAB

Semidefinite programming solver

In YALMIP

LMILAB is invoked by using 'lmilab' as solver flag in sdpsettings

Comments

Summary: Please do not use LMI Lab with YALMIP. Install another semidefinite programming solver such as MOSEK, SEDUMI or SDPT3

Please note that LMI Lab often is terribly slow on general problems. LMI Lab is efficient when control specific structure is explicitly stated, but YALMIP cannot exploit this feature and LMI Lab will thus run much slower compared to when the problem is modeled directly in LMI Lab. Hence, if you insist on using LMI Lab, YALMIP will cause efficiency problems. Of course, a simple solution is to switch solver and use just about any other solver.

Another problem with the solver is that LMI Lab does not return duals.

Additionally, LMI Lab does not return any error or infeasibility reports to YALMIP. Hence, YALMIP will always report success, no matter what happens.

Finally, LMI Lab has a very low precision tolerance by default, so you often get solutions far from optimal.

To summarize, do not use LMI Lab together with YALMIP unless you really, really, have to. And then, if you do, you are on your own.

Leave a Comment