hello, i've moved to @notwa@cybre.space. you should be receiving a follow shortly~
hello, i've moved to @notwa@cybre.space. you should be receiving a follow shortly~
ad-hoc optimizer i'm toying with. convergence is poor, but it works as a proof of concept (mloss is accuracy%) https://mastodon.social/media/u8LbvGapq9xJb_3ea6o
optimizer rambling
ok I think Gray code ES is kinda cheating with test functions with minima on a grid like Rastrigin, because the mutations tend to only move one dimension at a time. it would be more difficult if the minima weren't so evenly spaced. then again, maybe this goes for gaussian noise mutations too though, I dunno, it just doesn't seem right to me. nonetheless, it still performs relatively well. I kinda wanna try it on a stochastic problem next; I feel like its greediness will fail.
proof of concept for now. i'm sure this has been done before but whatever, i'm having fun https://mastodon.social/media/9xGfH97VMOd2xNrxRq0
boring boy
https://eaguru.guru/