📢 The workshop on Scientific Methods for Understanding Deep Learning is happening on Sunday in West meeting rooms 205-207 at #NeurIPS2024! We can't wait for Sunday! 🤩 Useful links below ⬇️

NeurReps 2025 is starting to take shape! If you're interested in serving in the Program Committee take a look at this form 👇

NeurReps is back for its 4th edition at NeurIPS 2025!

Excited to co-organize our NeurIPS 2025 workshop on Foundation Models for the Brain and Body! We welcome work across ML, neuroscience, and biosignals — from new approaches to large-scale models. Submit your paper or demo! 🧠💪

We are back for a third edition !!! 🔵🔴

Why do video models handle motion so poorly? It might be lack of motion equivariance. Very excited to introduce: Flow Equivariant RNNs (FERNNs), the first sequence models to respect symmetries over time. Paper: arxiv.org/abs/2507.14793 Blog: kempnerinstitute.harvard.edu/research/deepe… 1/🧵

NeurReps is back for a fourth edition! See you in San Diego to discuss the intersection of geometry, topology, representation learning, mechanistic interpretability, and neuroscience

🚨 The NeurReps 2025 Call for Papers is now OPEN! 🚨 Submit your contributions by August 22.

🚨 The UniReps 2025 Call for Papers is now OPEN! 🚨Submit your contributions by August 22.

Come join us in San Diego this year!

🚀 New Open-Source Release! PyTorchTNN 🚀 A PyTorch package for building biologically-plausible temporal neural networks (TNNs)—unrolling neural network computation layer-by-layer through time, inspired by cortical processing. PyTorchTNN naturally integrates into the

NeurIPS Conference, why take the option to provide figures in the rebuttals away from the authors during the rebuttal period? Grounding the discussion in hard evidential data (like plots) makes resolving disagreements much easier for both the authors and the reviewers. Left: NeurIPS

An introduction to information geometry in 10 pages: "The Many Faces of Information Geometry" Open access PDF

I derived a category-theoretic notion of a (CUDA) kernel as a parallelised function that works *shockingly* well, turning fusion into a compositional property. The remaining hurdle is figuring out how to deal with streamable/looped operations.

🚨 The submission deadline is approaching! 🚨 Don't miss the chance to share your work at NeurReps (NeurIPS Conference 2025). Submit by August 22 🧠

🎉 Big news! 🚨 The NeurReps Prize offers $500 for the best paper in these categories: 🧠 Neuroscience + Interpretability 🤖 Topological + Geometric ML ❄️ Symmetry + Equivariance Submit by Aug 22 for your chance to win! And thanks to New Theory for sponsoring! 🙌

📢 #NeurReps2025 UPDATE While we are beginning to set up the reviewing process, we're accepting late submissions until September 4th! 📅 Get your papers in now to be part of this exciting #NeurIPS2025 workshop on neural representations. Submit at neurreps.org/call-for-papers

1/2) I am very happy to finally share something I have been working on and off for the past year: "The Information Dynamics of Generative Diffusion" This paper connects the entropy production, divergence of vector fields and spontaneous symmetry breaking in a unified framework

Due to a record # of submissions for our NeurIPS workshop on Symmetry & Geometry in Neural Representations (neurreps.org), we're looking for volunteers to serve as Reviewers. So many phenomenal submissions. Send me your OpenReview email if you'd like to contribute!

Our #DynamicalSystems #FoundationModel was accepted to #NeurIPS2025 with outstanding reviews (6555) – first model which can *0-shot*, w/o any fine-tuning, forecast the *long-term statistics* of time series provided a context. Test it on #HuggingFace: huggingface.co/spaces/Durstew… ...