I've never managed to produce a useful function using GP but have found that in some instances, GAs can outperform both simulated annealing and the great deluge

I've never managed to produce a useful function using GP but have found that in some instances, GAs can outperform both simulated annealing and the great deluge algorithm. Adding random mutations to the process can be especially useful when there are local minima/maxima that would stop an e.g. GDA instead of finding the global minima/maxima.