Watchmaker's GA implementation is essentially unmaintained. If you want an efficient evolutionary algorithm implementation that *is* maintained, you might want to check out org.apache.mahout.ep.EvolutionaryProcess which is the recorded step meta-mutation implementation that drives the tuning of the learning knobs on the SGD classifiers. This class does not do a traditional genetic algorithm and prefers to optimize functions of continuous parameters, but it is many orders of magnitude more efficient.
See http://arxiv.org/abs/0803.3838 for details on this algorithm. On Tue, Apr 19, 2011 at 9:19 AM, Wojciech Indyk <[email protected]>wrote: > > Is this for production optimization? Or for self-education? > It's for self education and probably research (on university), because > it's very intresting project, but for example to research efficiency > of watchmaker in cloud i must launch it first :)
