RT @BachFrancis: My book is (at last) out, just in time for Christmas! A blog post to celebrate and present it: htt…

Come say hi at the noon poster session today in the East hall, poster #2201. 🚀 #NeurIPS2024

@AggieInCA Thank you for the kind words, Vimal. :)

Might be of interest to @ylecun @randall_balestr @imisra_ @tydsh @garridoq_ @AggieInCA @jsusskin

RT @OdedRechavi: Scared to share this after 10 freaking years of waiting but here we go: We found that temperature *perception* can change…

@naturecomputes ✋I will be there too. :) Would be fun to catch up. :)

@sarthak__saha @perrys_lab Congratulations!! 🥳 When do you move to Boston? :)

@Ahmed_Masry97 @ArnabMondal96 Yup, that's correct! Also, Canada recently made changes to the PR process, as of Nov 2024. 🥲

@roydanroy That's not quite true, is it? (Depends on the province right?)

RT @RichardGerum: 🎉 Great news! Pylustrator, my open-source tool for creating publication-ready @matplotlib plots, works seamlessly with th…

RT @k47h4: I am excited to announce that I've been awarded an SNSF Starting Grant! 🥳 I will start a lab at the Institute of Neuroinformatic…

@dlevenstein My guess would be that we need to have particular initializations that favor/maintain these structures over training, but our current initialization strategies may not be the ones. One example that comes to mind is sparsely connected models that require a diff init setup.