[Koffiepoeder]

Hey Justin & Nick,

Recently I have been experimenting a lot with optimization strategies and their performance/results. I have tried to code some, including mixed integer LP, and applied them to multiple problems. I did this such that I could get a deeper understanding which sort of optimization algorithm is suited and well-performing for which kind of problem.

So far my results have been very inconsistent: sometimes genetic algorithms produce surprisingly good results, other times simulated annealing proves to be the clear winner, sometimes a plain old depth-first search, with a custom scoring heuristic produces greatest results... Since my results seem quite random to me so far, I get the impression that the only way to find out the best match for a problem is to try them all. Is this correct?

Since you're using mixed integer programming, could I ask you why you chose this optimization strategy? Also, if you would happen to have some more information about the specifics of choosing a correct algorithm for an optimization problem, could you provide me some insights?

Anyways, good luck with Coursedog!

[jwenig]

I wish I could tell you that I was an expert, but the truth is that we worked on a slow as 1990's software genetic algorithm before we found success with the MIP. :-) If you shoot me an email [e-mail] I can connect you with our algorithms engineers who have the background to discuss