Replies

Along those lines, I wrote a short simulation to test the effects of mutation and selection. (Not part of a project, just a personal thing at home.) I set up an array of individuals (about 50000) each with an array of several (about 5) integers to represent "genes." The "genes" underwent (sometimes biased) Brownian motion by adding or subtracting numbers based on a PRNG. The "phenotype" of an individual was computed (partly randomly) from the gene set. A "fitness" function was applied to the genotypes to give a reproduction rate. Then a new generation is created with Russian Roulette to delete the "less fit."

One result was that choosing a bias to smaller integers for genes in the mutation phase and choosing a fitness as having larger integers in the genes resulted in a population of large genes.

The selection phase seemed to work proportional to time and the muataion phase proportional to the square root of the time. This seems to be the correct orders of magnitude.