Training Neural SDEs: We worked out how to do scalable reverse-mode autodiff for stochastic differential equations. This lets us fit SDEs defined by neural nets with black-box adaptive higher-order solvers. arxiv.org/pdf/2001.01328… With Xuechen Li, @rtqichen and Leonard Wong.

Excited to share our work on "Implicit Neural Representations with Periodic Activations" vsitzmann.github.io/siren We show how to fit complex signals, such as room-scale SDFs, video, & audio, and supervise implicit reps via their gradients to solve boundary value problems! (1/n)

[1/5] *torchdyn* goes generative! This new release adds continuous normalizing flows Ricky T. Q. Chen will grathwohl David Duvenaud , energy-based Neural DEs Miles Cranmer Sam Greydanus, and extends existing support for higher-order variants. All powered by Lightning! @_willfalcon

We just open-sourced differentiable SDE solvers in PyTorch: github.com/google-researc… Now you can put stochastic differential equations in your deep learning models, and neural nets in your SDEs! Credit to Xuechen Li.

Work done with wonderful co-authors: Josh Romoff, David Kanaa Emmanuel Bengio atouati Pierre-Luc Bacon and Joelle Pineau!!

His fight is over. May the spirit of the Black Panther rest in peace in the ancestral plane. 🙏🏾

I’m having an especially hard time concentrating on research this week, but instead of doomscrolling on twitter I guess I’ll do something semi-productive and make another twitter thread. Today's topic: What is a Lyapunov function? 1/

Anyone else?

Last apple picking day with my group at Mila - Institut québécois d'IA. I'm grateful to be able to work with such talented and kind researchers! 🍎

New paper! 📄 "Continuous-Time Meta-Learning with Forward Mode Differentiation", with David Kanaa, Leo Feng, Giancarlo Kerg, Yoshua Bengio, Guillaume Lajoie & Pierre-Luc Bacon Mila - Institut québécois d'IA Accepted as a Spotlight at #ICLR2022 paper: arxiv.org/abs/2203.01443 code: github.com/tristandeleu/j…

Decision-aware RL (aka control-oriented sysid in other fields) has drawn a lot of attention lately in model-based RL. We're happy to host the first workshop on this topic at ICML Conference. An occasion to collectively reflect on the successes, challenges and path forward in this field

ICLR in Africa for the very 1st time! O my, o my! 🤩🗓️🇷🇼📍

Course Correcting Koopman Representations Accepted at #ICLR2024! We identify problems with unrolling in imagination and propose an unconventional, simple, yet effective solution: periodically "𝒓𝒆𝒆𝒏𝒄𝒐𝒅𝒊𝒏𝒈" the latent. 📄 arxiv.org/abs/2310.15386 Google DeepMind 1/🧵

The Black @ Mila Collective is excited to host the Tea Talk series on Fridays, from February 2nd to 23rd at 10:30 am. We have a notable line-up of Black professionals who will each give a one-hour scientific presentation about their expertise in AI research. Tune in! #AI #BHM

Last week, I gave a talk at Mila - Institut québécois d'IA. The talk should be of interest to anyone working on predictive models, particularly in latent space. In collab. with Mahan Fathi Clement Gehring Jonathan Pilault David Kanaa Pierre-Luc Bacon. See you at ICLR 2026 in 🇦🇹! drive.google.com/file/d/1mQSXFa…