@gradientflowing.bsky.social
Keep the gradients flowing
Look how awkward I am in front of a camera π€¦πΌββοΈ But the science we do is great, I swear, get in touch if you'd like more details!
10.04.2025 15:07 β π 6 π 1 π¬ 0 π 0What do we need, in good as in bad times? A community. Today is as good a day as any to:
β¨renew your membership to the Society for the Neurobiology of Language
β¨submit an abstract/symposium to #SNL2025
β¨ register for one of the member-initiated virtual activities 2025.neurolang.org/virtual-acti...
Missing from Martin's great recap of #ABIM2025 is his own talk which covered in 30 min ins&outs of representational spaces in behavior, brains, & artificial agents. This edition of ABIM was all-around terrific: check out what you missed (www.unige.ch/ABIM/files/9...) & set a reminder for next year!
16.01.2025 13:47 β π 9 π 2 π¬ 2 π 0And our daughter too. almost forgot about that π
02.01.2025 16:09 β π 1 π 0 π¬ 0 π 0Sport was indeed the best part of 2024.
01.01.2025 08:17 β π 6 π 1 π¬ 1 π 0When optimization problems have multiple minima, algorithms favor specific solutions due to their implicit bias. For ordinary least squares (OLS), gradient descent inherently converges to the minimal norm solution among all possible solutions. https://fa.bianp.net/blog/2022/implicit-bias-regression/
19.12.2024 06:00 β π 24 π 5 π¬ 2 π 0
Finishing a very sporty 2024 in the best possible way #sansilvestre2024 @vborghesani.bsky.social
31.12.2024 18:00 β π 3 π 0 π¬ 0 π 1Very nice post from my colleague Rishi Mehta on the different tactics used by AlphaProof during the international math olympiad
rishimehta.xyz/2024/11/17/a...
New essay by DeepMind about AI for scientific discovery, there's a lot of interesting ideas and citations to others's work here
deepmind.google/public-polic...
