Nice start to 2025 - finally bullied a toy reinforcement learning model I've been using to try stuff out into different behaviours for different parameters.
It only appears in the speed to converge, in the early learning stage. I was expecting to get a more qualitative difference where the model can just fail within certain parameter ranges but no luck there yet.
[#]actorcritic #AI
=> More informations about this toot | More toots from TEG@mastodon.online
text/geminiThis content has been proxied by September (ba2dc).