🔥The 9th Round of Easy Loan, Earn $40 Reward is in progress❗️ ⏰ Promotion Period: January 15th - Feburary 15th, 2025 👉 Register now and check more details at gate.io/campaigns/358

Blog post: Rethinking Probability - Mass, Averages, and Granularity Developing an intuition for probability using analogies from physics, exploring both the standard measure-theoretic and the expectation-first foundations.

see you all in Vancouver 🇨🇦 new results on PDE benchmarks soon as well 👨‍🍳

#compchem Second preprint linked to the FeNNix-Bio1 #machinelearning foundation model. FeNNix-Bio1's inference is pretty fast already with a few GPUs but, "what if", we were able to push it at the #Exascale? Let's have a glimpse into the future (1/3): "Pushing the Accuracy Limit

After supervising 20+ papers, I have highly opinionated views on writing great ML papers. When I entered the field I found this all frustratingly opaque So I wrote a guide on turning research into high-quality papers with scientific integrity! Hopefully still useful for NeurIPS

LLMs are a classic example of emergence, and it is not the first time we look into a non-living system showing life-like properties, Prigogine got a Nobel for that. I totally see Anthropic going this direction and getting another one.

We released the Open Molecules 2025 (OMol25) Dataset last week! 🚀🧪 OMol25 is a large (100M+) and diverse molecular DFT dataset for training machine learning models. It was a massive collaborative and interdisciplinary effort and I’m super proud of the whole team! 🙌 1/7

The arxiv preprint on our conformally equivariant neural network -- named AdS-GNN due to its secret origins in AdS/CFT -- is now out! arxiv.org/abs/2505.12880 🧵explaining it below. Joint work with the amazing team of Max Zhdanov, Erik Bekkers and Patrick Forre.

New preprint! We extend Taco Cohen's theory of equivariant CNNs on homogeneous spaces to the non-linear setting. Beyond convolutions, this covers equivariant attention, implicit kernel MLPs and more general message passing layers. More details in Oscar Carlsson's thread 👇

1. We often observe power laws between loss and compute: loss = a * flops ^ b + c 2. Models are rapidly becoming more efficient, i.e. use less compute to reach the same loss But: which innovations actually change the exponent in the power law (b) vs change only the constant (a)?

Interested in graph and topological deep learning?🍩 Join us this wednesday online for this exciting Symmetry and Geometry in Neural Representations global seminar series!🌟

Excited to be giving a talk at the Cambridge Wednesday Seminar today at 3pm. Looking forward to sharing ideas and great discussion about equivariance and beyond Thanks Pietro Lio' Riccardo Ali for inviting me! cst.cam.ac.uk/seminars/list/…

ICML Spotlight 🚨 Equivariance is too slow and expensive, especially when you need global context. It makes us wonder if it even worths the cost, especially in high-dimensional problems? We present Geometric Hyena Networks — a simple equivariant model orders of magnitude more

Neural networks implicitly define a latent vector field on the data manifold, via autoencoding iterations🌀 This representation retains properties of the model, revealing memorization and generalization regimes, and characterizing distribution shifts 📜: arxiv.org/abs/2505.22785

Great discussion, Chaitanya K. Joshi! We also explored this with extensive experiments in our recent paper: arxiv.org/abs/2501.01999. We find, among others, that equiv mods in a sense scale even better than non-equiv ones. Going more or less completely against the vibes from your post😅1/5