Teddy retweeted
🧵0/7
🚨 Spotlight @ICML2025 🚨
Chi-Ning and Hang have been thinking deeply about how feature learning reshapes neural manifolds, and what that tells us about generalization and inductive bias in brains and machines.
They put together the thread below, which I’m sharing on their behalf 👇Enjoy!
Teddy retweeted
In many brain areas, neuronal tuning is heterogeneous. How does this diversity help behavior? We show how tuning diversity shapes representational geometry and boosts coding efficiency for perception in our new preprint: biorxiv.org/content/10.1101/…
(with @s_y_chung & Tony Movshon)
Very fun project providing a formal link between MMCR and information theory!
🔥🚀Towards an Improved Understanding and Utilization of Maximum Manifold Capacity Representations🚀🔥
@vclecomte @BerivanISIK @dhruv31415 @carranzamoreno9 @AlyssaUnell @KhonaMikail @tedyerxa
& @sanmikoyejo @ziv_ravid @Andr3yGR @s_y_chung @ylecun
arxiv.org/abs/2406.09366
1/9
Excited to share our results on Efficient Coding of Natural Images using Maximum Manifold Capacity Representations, a collaboration with @KuangYilun @EeroSimoncelli and @s_y_chung to be presented at #NeurIPS2023
1/n
Check out the full paper for more fun, including some empirical results suggesting a mechanism by which augmentation based SSL leads to semantic clustering!
Paper: openreview.net/pdf?id=og9V7N…
Code: github.com/ThomasYerxa/mmc
17/n
Also, be sure to check out interesting followup work that formalizes a connection between our MMCR loss function and information theory, from @RylanSchaeffer & colleagues: xitter.gq/RylanSchaeffer/status/…
18/n, n=18
Excited to begin announcing our #NeurIPS2023 workshop & conference papers (1/10)!
🔥🚀An Information-Theoretic Understanding of Maximum Manifold Capacity Representations🚀🔥
w/ amazing cast @vclecomte @BerivanISIK @sanmikoyejo @ziv_ravid @Andr3yGR @KhonaMikail @ylecun
1/7
