On Aug 8, 2012, at 10:17 AM, erwinski wrote: > Are there any plans to add Particle Swarm Optimization (PSO) or other > metaheuristic algorithms such as Differential Evolution to NLOPT? I think > this could be useful to some people especially when combined with AUGLAG. > What should I do to add an additional algorithm to NLOPT in order to compile > it using the current compile system (using configure/make/make install)? I've > implemented PSO in NLOPT on my own but did it in a rather unelegant and > probably unoptimal way by adding all nlopt C-files (with some modifications) > to my eclipse project and compiling everything into one executable.
I hadn't planned on adding PSO, as whenever I've tried PSO in the past it had substantially worse convergence rates on all of my test problems compared to the other algorithms already in NLopt. Have you had better experiences with your implementation? If you have a free-software PSO code ready for inclusion in NLopt, I'd certainly consider including it. The ISRES algorithm uses differential evolution. > Another suggestion would be to change the "double* work" pointer in the > "nlopt_opt" struct (pointer to algorithm specific workspace) into a void > pointer so one could allocate an arbitrary workspace like a structure not > just an array of doubles. Will do. This is included in the next version, to be out soon due to a bugfix I just noticed. Steven _______________________________________________ NLopt-discuss mailing list [email protected] http://ab-initio.mit.edu/cgi-bin/mailman/listinfo/nlopt-discuss
