PROC OPTMODEL: NLPC Vs. NLPU

[vioravis]

I am using PROC OPTMODEL for optimizing a nonlinear function created using PROC FCMP. The function to be optimized is a complex nonlinear function with two variables and there are only bound constraints on the two variables (no other constraints)

There are 3 solvers available: NLPU, NLPC and SQP. First one seems to be for unconstrained problems but it does allow for bound constraints. In my case, it appears that I can use any one of the 3 solvers since only bound constraints are there. I have tried (70+ test cases) with NLPC and NLPU. NLPC seems 3-4 time faster than the NLPU. However, there are cases where NLPC crashes but NLPU works and vice versa.

Are there any guidelines on when to use/when not to use these solvers? Is there a way to debug to find out why one solver is working while the other crashes.

Thank you.

Ravi

[RobPratt]

As of SAS 9.3 released in July 2011, the NLPU, NLPC, SQP, and IPNLP solvers are no longer under active development.  You should instead use the NLP solver (SOLVE WITH NLP), which accesses two algorithms: INTERIORPOINT (default) and ACTIVESET.  For more details, see:

http://support.sas.com/documentation/cdl/en/ormpug/63975/HTML/default/viewer.htm#ormpug_nlpsolver_se...

Also, FCMP support for PROC OPTMODEL is undocumented and not yet fully functional.