9 ms·
Annealing is hit or miss. I often have to sample a distribution on a domain that approximates an infinite dimensional space and annealing doesn't cut it for me
by coherentpony 14y ago
Annealing is hit or miss. I often have to sample a distribution on a domain that approximates an infinite dimensional space and annealing doesn't cut it for me. There are just far too many modes. Personally, I'd like to see what GAs and GPs have to offer in this regard.
- jacques_chester 14y agoWell to pick between GA and GP, the question is: are you trying to evolve a list of parameters, or create a novel function/program? Parameters into function: use GA. Function/program: use GP. (speaking very roughly, that's the historical distinction between them)
- coherentpony 14y agoFrom a mathematical perspective, it looks as though your statement could imply that GA may potentially be viewed as a finite dimensional analogue of GP. Interesting.