#KalmanFilters are probably among the most important/useful topics that I missed studying during my PhD, despite all the months of lecture on approximation theory that I sat in and did homeworks for. Trying to catch up now! #AI4Science

To the best of my surveying so far, the finite-sample guarantee for #KalmanFilter is fairly recent - surprisingly so! It's this Ghattas-Alonso paper doi.org/10.1093/imaiai… at #Information&Inference - one of my favorite journals for obvious reasons 😉

I got selected for the "Early Career Highlights" of ACM IKDD International Conference on Data Science (#CODS) 2025. Looking forward to giving the related presentation in December at IISER, Pune. acm.org/articles/acm-i…

Today is 70th anniversary of the summer meeting at Dartmouth which officially marked the beginning of AI research 💥 Interestingly "Objective 3" in 1955 was already about having theory of neural nets. 🙂 stanford.io/2WJJJGN

If you ever needed a paper to start reading LLMs from, then it's hard to find anything better than what's published at venues like these. Do read these by one of my former advisers ! 🔥

Recently I gave an online talk IISc Bangalore 's "Bangalore Theory Seminars" where I explained our results on size lowerbounds for neural models of solving PDEs via neural nets. #SciML I cover work by my 1st year student, Sebastien. youtu.be/CWvnhv1nMRY?fe…

Registrations close for #DRSciML by Noon (Manchester time). Do register soon to ensure you get the Zoom links to attend this exciting event on the foundations of #ScientificML 💥

Do register here, https:// drsciml.github.io/drsciml/ 🔥

At #DRSciML, my student Dibyakanti Kumar presenting our results on sample complexity of solving Navier-Stokes PDE via neural nets. Possibly the first such generalization error bounds for a non-linear PDE. #AI4Science #ScientificML

With all the discussion about #DeepMind 's use of PINNs to "find" unstable singularities in 3D Euler PDE, I feel like advertising that more than a year ago, we had shown that generalization bounds exist for these setups. iopscience.iop.org/article/10.108…

With all the discussion about "Sparse AutoEncoders" as a way of doing #MechanisticInterpretability of LLMs, I am resharing a part of my PhD where we proved years ago about how sparsity automatically emerges in autoencoding. Neel Nanda arxiv.org/abs/1708.03735

In particular, in this work we identified how in an infinitely wide limit, the autoencoding landscape starts to encode the true "data dictionary" - a fully non-realizable data setup. We were kind-of close to discovering the NTK idea - but we missed it 🫥

Totally! There is an outsized effect of strong memory on ability to do high-end mathematics.

With all the discussions about #Muon and "#silversteps" etc, here's my periodic reminder that we invented delta-GClip - the *only* step-size schedule for GD that provably trains deep-nets. This is a logical benchmark for all deep-net training claims 💥

I am ready to bet, that OpenAI's historic successes can be traced back to Ilya Sutskever taking differential topology classes as an undergrad. 😁 Seriously, it's not inconceivable that one of the greatest minds in ML history also found G-bundles interesting!

I had thought I will keep my X feed fully academic. But this sporting victory by Indian women is just too big. 2nd Nov 2025 goes down in history 🔥 #CWC25