6. "Our Genetic Algorithm mutate each solution, we do not use any probability to determine chance of mutation occurence like in a typicial GP." - Ok, whatever works for you, but in my opinion you shouldn't do it like that. Read about Eigen's threshold, if mutations occur too often, the system will be incapable of retaining any information (in this case - best solutions). Think about it, if you find an almost perfect solution, it won't pass its fitness much further, because it will be destroyed in a few generations.
jkarolczak opened this issue · comments
Zgłębiłem temat i rzeczywiście ma to sens. Do tego zmieniłem odrobinę heurystykę, aby także w tym miejscu uniknąć utykania w lokalnych optimach