Improve default value for mutation rate in BinaryGenAlgSolver

[mikeliux]

In BinaryGenAlgSolver, the current default value for mutation rate might be too large for some problems (i.e., where a chromosome has hundreds of genes), to the point that the algorithm degenerates into a random search and cannot find optimum values within 1000 generations.

If a fixed mutation rate is going to be used and the user defines no value for this parameter, then the simplest and best initial guess would be:

self.mutation_rate= 1/self.n_genes

More info available here by MIT press

I actually produced a heatmap with different mutation_rate and selection_rate values for a couple of test functions, by computing the average minimum value reached after a fixed number of iterations. The best value for mutation_rate in this experiment was almost exactly 1/n_genes, which agrees with the recommendation given in the literature. I found no best value for 'selection_rate', most values yielded similar results.

[diogomatoschaves]

Hi @mikeliux,

Yeah that sounds like a good default value for mutation_rate! Do you want to submit a PR for that change?