I've added a small tweak to selDoubleTournament which allows adding elitism to the size tournament, so that parsimony pressure only applies when fitnesses are equal. The default behaviour (fitness_greedy=False) is the same as the old behaviour, since it may not be helpful in all circumstances.
I've found it converges much quicker and more reliably on a minimisation problem with an integer-valued fitness function. It may be terrible with continuous fitnesses where a tie is rare. Perhaps it would be better to use an epsilon parameter and let the size tournament behave as normal unless the fitnesses are within some distance d of each other? I'd be interested to hear your thoughts.
I've added a small tweak to
selDoubleTournament
which allows adding elitism to the size tournament, so that parsimony pressure only applies when fitnesses are equal. The default behaviour (fitness_greedy=False) is the same as the old behaviour, since it may not be helpful in all circumstances.I've found it converges much quicker and more reliably on a minimisation problem with an integer-valued fitness function. It may be terrible with continuous fitnesses where a tie is rare. Perhaps it would be better to use an epsilon parameter and let the size tournament behave as normal unless the fitnesses are within some distance
d
of each other? I'd be interested to hear your thoughts.