Efficiency of crossover in genetic algorithms -


i've implemented number of genetic algorithms solve variety of problems. i'm still skeptical of usefulness of crossover/recombination.

i first implement mutation before implementing crossover. , after implement crossover, don't typically see significant speed-up in rate @ candidate solution generated compared using mutation , introducing few random individuals in each generation ensure genetic .

of course, may attributed poor choices of crossover function and/or probabilities, i'd concrete explanation/evidence why/whether or not crossover improves gas. have there been studies regarding this?

i understand reasoning behind it: crossover allows strengths of 2 individuals combined 1 individual. me that's saying can mate scientist , jaguar smart , fast hybrid.

edit: in mcdowella's answer, mentioned how finding case cross-over can improve upon hill-climbing multiple start points non-trivial. elaborate upon point?

you correct in being skeptical cross-over operation. there paper called "on effectiveness of crossover in simulated evolutionary optimization" (fogel , stayton, biosystems 1994). available free @ 1 (a source of lot of other great pubs fogel well).

by way, if haven't recommend looking technique called "differential evolution". can @ solving many optimization problems.


Comments

Popular posts from this blog

c# - How to set Z index when using WPF DrawingContext? -

razor - Is this a bug in WebMatrix PageData? -

visual c++ - Using relative values in array sorting ( asm ) -