This thesis contains experimental evaluation of results of Livnat et al. (Satisfiability and evolution, FOCS, 2014), who have shown that, given a logical function, the population of binary vectors converges to a population of function models, under the assumption of weak selection and product reproduction. Our experiments worked with population size 10000 and genotype length 28. We have experimentally confirmed their theoretical results and for a handful of special logical functions shown that the convergence speed towards the population of models matches the convergence speed of monotone logical functions.
We have also run the experiments using crossover and mutations reproduction model. In this case we have for a random logical function experimentally detected convergence towards a population with high ratio of models. However, there are some classes of logical function for which we failed to achieve a similar convergence.
|
