5 ms·
From personal experience, I've always had better results from Simulated Annealing than Genetic Programming. I hasten to point out that it's entirely possible I'
by strags 6y ago
From personal experience, I've always had better results from Simulated Annealing than Genetic Programming. I hasten to point out that it's entirely possible I've been doing GP poorly - but part of the problem seems to me that GP has so many more knobs/dials/parameters that need to be tweaked correctly in order to yield good results.
- pixelpoet 6y agoDo you have any good references for SA? Thanks!
- red2awn 6y agoJust read wikipedia. [0] is also a good high-level overview of the different algorithms in metaheuristics. [0]: https://cs.gmu.edu/~sean/book/metaheuristics/Essentials.pdf https://cs.gmu.edu/~sean/book/metaheuristics/Essentials.pdf
- joe_the_user 6y agoResults on what? The effectiveness of an algorithm depends on the problem you target. Edit: Also, if you broaden the definition just a bit, you could call simulated annealing a kind of genetic program.
- nextos 6y agoAre you referring to GA or GP? GP is not comparable to simulated annealing, as it's kind of non-parametric. Many ideas of GP are getting reused in Bayesian program induction, in conjunction with differentiable programming, SAT solvers, etc [1]. IMHO a very promising route to AGI. [1] https://web.mit.edu/ellisk/www/documents/dreamcoder_with_supplement.pdf https://web.mit.edu/ellisk/www/documents/dreamcoder_with_sup...
- SeanLuke 6y agoThese are orthogonal things. Simulated annealing is an algorithm. Genetic Programming is pursuit, which can (among other things) use simulated annealing to achieve its goal. Are you sure you're not confusing this with something else? Did you mean Koza-style GP perhaps? Or did you mean a genetic algorithm (GA)?