Re:The Genetic Algorithm Explained
by Houshalter on Tuesday February 17, 2015 @09:54PM
(#9066748)
In practice genetic algorithms don't do much better than hill climbing with random restarts.
You can do random restarts, stochasticism, or population based stuff on gradient methods too.
Yes it does require at least an estimate of the gradient, but that's not too hard to get and it converges much faster. Meta-heuristics are only a last resort when you can't get a gradient.
Although they are a bit simpler so there is nothing wrong with using them if the search space isn't too big.
You can do random restarts, stochasticism, or population based stuff on gradient methods too.
Yes it does require at least an estimate of the gradient, but that's not too hard to get and it converges much faster. Meta-heuristics are only a last resort when you can't get a gradient.
Although they are a bit simpler so there is nothing wrong with using them if the search space isn't too big.