Conversation

Notices

1. Embed this notice
   iced depresso (icedquinn@blob.cat)'s status on Monday, 28-Oct-2024 04:15:56 JST iced depresso
   funny to see one research line talk about the neocortex only having about six layers--a different unrelated group stumbled on a similar theory that only a small amount of the brain participates in learning.
   
   the biologically inspired models share issues that they work great but do not scale vertically. :blobcatovo: i somehow knew that we'd learn more by studying the limitations and amusingly we're getting more research that deep learning is a meme and the algorithms break down in weird ways
   
   not sure how to reconcile the self organizing maps though. there seems to be a weird amount of success in putting a (large) self organizing map in front and a small learning layer behind it.
   
   :neocat_thonk: still somewhat curious if KANs are worth trying to mix with a spike net. i have no real theory why we should do this. it just seems like something that should be tried.
   • Embed this notice
     Doughnut Lollipop 【記録係】:blobfoxgooglymlem: (tk@bbs.kawa-kun.com)'s status on Monday, 28-Oct-2024 04:19:07 JST Doughnut Lollipop 【記録係】:blobfoxgooglymlem:

     in reply to
     @icedquinn Neo Cortex has many layers. :blobfoxgooglymlem:
   • Embed this notice
     iced depresso (icedquinn@blob.cat)'s status on Monday, 28-Oct-2024 04:29:09 JST iced depresso

     in reply to

     • Doughnut Lollipop 【記録係】:blobfoxgooglymlem:
     @tk a few :blobcatovo:
     
     its more wide than tall
     Doughnut Lollipop 【記録係】:blobfoxgooglymlem: likes this.
   • Embed this notice
     iced depresso (icedquinn@blob.cat)'s status on Monday, 28-Oct-2024 04:35:33 JST iced depresso

     in reply to

     • THOT POLICE
     @s8n it primarily happens in a small part of the neocortex (the hippocampus.)
     
     there are some weird properties with learning rules though. had a paper on friday where they were getting 95% on MNIST purely through plasticity rules and no actual learning engine.
     
     there seems to be some amount of self-organization with a feedback governor of some kind. but we know mammals can't learn shit when the hippocampus is broken, so that seems to be where the feedback governance is being emitted from.
     
     (the prefrontal sends a lot of stop signals though.)
   • Embed this notice
     THOT POLICE (s8n@posting.lolicon.rocks)'s status on Monday, 28-Oct-2024 04:35:35 JST THOT POLICE

     in reply to
     @icedquinn learning is probably governed by the most complex parts of the mind, the parts that get angry when contradicted etc