Hi, I am new to nlopt (and haven't done any optimization for years and years...) I also feel this is probably a FAQ, but I've had a bit of a search and cannot find anything...
I am doing all this in C++ but using the C opt interface as the C++ interface has some incompatibilities with other libraries my framework obliges me to use... For the purposes of evaluating my objective function, I would like to display the evolving parameter set at intermediate points through the optimization process. Can I do this by:- setting maxeval to a moderate number- when that terminates, take the current parameters and display them- then optimize again using the current parameters as a new start point... The thing against this is that second and subsequent runs have presumably lost any working state that was created in previous runs, so my precise questions are: 1) can I optimize again on the same nlopt_opt object (I assume I can) 2) if I do so will I suffer from "starting again" each time? how much might this impact performance? 3) is there a way to tell nlopt to "resume" rather than "start over"? 4) or alternately can I pull some working state out of the optimizer after one run and reinject it before the next(e.g. I have access to initial step sizes, so can I read out final step sizes and supply those to the next run, and there any point to this if there is other internal state than I cannot preserve?) I do not know exactly what algorithm I will be using, but I intend to constrain the objective to forms I can easily evaluate gradients for, and a local minimum suited to my purposes (I'm designing the system under optimization and have some flexibility to try and design poorly-optimized minima out of it...) Thanks in anticipation, Ian
_______________________________________________ NLopt-discuss mailing list NLopt-discuss@ab-initio.mit.edu http://ab-initio.mit.edu/cgi-bin/mailman/listinfo/nlopt-discuss
