4 ms·
It would seem intuitive that the search process would be greatly speeded up by using a genetic search of the solution space rather than brute force search, it's
by fchollet 14y ago
It would seem intuitive that the search process would be greatly speeded up by using a genetic search of the solution space rather than brute force search, it's a bit odd that genetic algorithms aren't mentioned.
- jacques_chester 14y agoTwo things. 1. You're thinking of genetic programming, not genetic algorithms. 2. They used the Markov chain Monte Carlo method to perform the search. It's a different method altogether. For example, note the absence of crossovers.
- Houshalter 14y agoI think that it would be correct to call it a genetic algorithm. I thought that genetic programming usually uses a tree structure in a functional programming language.
- jacques_chester 14y agoMost GP representations are of instruction trees, but some GP experiments have been done with strings of assembler instructions. The commonality is that the population of agents are programs and the test of fitness is a function execution.
- Houshalter 14y agoThe problem with that is that it's really hard to crossover strings of instructions in ways that don't completely break it, especially when code is inserted or removed from one parent. And crossover is 90% of evolution (in fact some claim that genetic programming works just as well or better with no mutations whatsoever and only crossover being used.)
- jacques_chester 14y agoIt's hard, but it's been done. The nice side effect is that fitness functions evaluate quickly.
- jules 14y agoIt isn't; genetic algorithms don't work any better than hill climbing.