SpeedUp of NSGA / SPEA
49 views
Skip to first unread message
Max Elfner
Jun 21, 2016, 4:15:51 PM6/21/16
to deap-users
Hi,
im using NSGA / SPEA optimization with a large number of individuals. I can use scoop to increase evaluation, but looking at pc usage shows that evaluating is not the main concern, but pareto-sorting. Anyone tried for a decent and "easy" way to speed up selection? Some stacked for-loops there which probably arent that optimal...
Cheers
im using NSGA / SPEA optimization with a large number of individuals. I can use scoop to increase evaluation, but looking at pc usage shows that evaluating is not the main concern, but pareto-sorting. Anyone tried for a decent and "easy" way to speed up selection? Some stacked for-loops there which probably arent that optimal...
Cheers
Marc-André Gardner
Jun 21, 2016, 4:21:28 PM6/21/16
to deap-users
Hi Max,
Did you try to use the sortLogNonDominated function (by passing nd='log' to the selNSGA2 function : https://deap.readthedocs.io/en/master/api/tools.html#deap.tools.selNSGA2 )? The complexity of this sort is quite lower than the canonical sortNonDominated and should help you get some decent speed up.
Marc-Andre
Did you try to use the sortLogNonDominated function (by passing nd='log' to the selNSGA2 function : https://deap.readthedocs.io/en/master/api/tools.html#deap.tools.selNSGA2 )? The complexity of this sort is quite lower than the canonical sortNonDominated and should help you get some decent speed up.
Marc-Andre
Max Elfner
Jun 22, 2016, 3:09:05 AM6/22/16
to deap-users
Did not yet, will try. Thanks for the advice.
Reply all
Reply to author
Forward
0 new messages