DiffeqML

Research group for the intersection of deep learning and dynamical systems.

Developers of *torchdyn*
Repo:

[DiffeqML @Diffeq_ml]

#ICML2022 week!📜

DiffeqML is contributing to the ICML22 AI4Science workshop:

1) Transform Once: Efficient Operator Learning in Frequency Domain

w/ Michael Poli, Stefano Massaroli, Federico Berto, Jinkyoo Park, Tri Dao, Christopher Ré, Stefano Ermon

When: July 23, 9.00AM EST.

[Michael Poli @MichaelPoli6]

❗New deadline for the #NeurIPS2022 workshop❗

"Symbiosis of Deep Learning and Differential Equations":

October 1st.

Website: https://t.co/WKy76ZMlLw. Send us your work on neural differential equations, learnable numerical methods, continuous-time diffusion and more!

[Michael Poli @MichaelPoli6]

Join us for the second edition of the #NeurIPS2022 workshop "The Symbiosis of Deep Learning and Differential Equations"🌀

We're looking for your AI <> DE ideas: neural diff. eqs., neural operators, diffusion models and novel applications!

website: https://t.co/O7VVNVLRX3

[DiffeqML @Diffeq_ml]

3) Unsupervised Discovery of Inertial-Fusion Plasma Physics using Differentiable Kinetic Simulations and a Maximum Entropy Loss Function

w/ Archis Joglekar, Alexander Thomas

All papers can be found at: https://t.co/RsDSj2brsp

[DiffeqML @Diffeq_ml]

#ICML2022 week!📜

DiffeqML is contributing to the ICML22 AI4Science workshop:

1) Transform Once: Efficient Operator Learning in Frequency Domain

w/ Michael Poli, Stefano Massaroli, Federico Berto, Jinkyoo Park, Tri Dao, Christopher Ré, Stefano Ermon

When: July 23, 9.00AM EST.

[DiffeqML @Diffeq_ml]

2) Efficient Continuous Spatio-Temporal Simulation with Graph Spline Networks

w/ Chuanbo Hua, Federico Berto, Michael Poli, Stefano Massaroli, Jinkyoo Park

Oral: July 23, 10.45AM EST.

[Michael Poli @MichaelPoli6]

Let us embark on a fractal journey about dynamical systems and neural implicit representations... 1/

[Stefano Massaroli @Massastrello]

Two papers accepted at #NeurIPS2021:
Differentiable Multiple Shooting Layers https://t.co/uwZd116rP9
Neural Hybrid Automata https://t.co/1nnWYCAzuw

@MichaelPoli6 @btreetaiji @animesh_garg

[Michael Poli @MichaelPoli6]

The website for our 'The Symbiosis of Deep Learning and Differential Equations' #NeurIPS2021 workshop is up: https://t.co/hU4sHcCqZD

We have a special track for already published papers. Share your work from adjacent fields with the NeurIPS community!

Deadline: Sept. 17 AoE

[DiffeqML @Diffeq_ml]

[1/6] Announcing **torchdyn version 1.0**: https://t.co/62Z9vnh9OT! @MichaelPoli6 @Massastrello.

We roughly doubled the number of tutorials (optimal control, parallel-in-time solvers, hybrid systems), added new models and developed a numerics suite for diff eqs and root finding

[DiffeqML @Diffeq_ml]

[2/2] "Hypersolvers: Toward Fast Continuous-Depth Models" accepted at @NeurIPSConf #NeurIPS2020.

Come chat with us:

Poster: Tue, Dec 8th, 2020 @ 09:00 – 11:00 PST (Session 2)

[DiffeqML @Diffeq_ml]

[1/2] NeurIPS is almost here!

"Dissecting Neural ODEs" accepted at @NeurIPSConf #NeurIPS2020 as oral.

Looking forward to your questions:

Oral: Tue, Dec 8th, 2020 @ 06:30 – 06:45 PST
Poster: Tue, Dec 8th, 2020 @ 09:00 – 11:00 PST (Session 2)
Code: https://t.co/pEzQB7ygDI

[Stefano Massaroli @Massastrello]

“Dissecting Neural ODEs” (#neurips2020 _oral_ paper) unveils the dynamical systems anatomy of continuous-depth learning from back-propagation to depth-varying parameters or state augmentation, while introducing several new models (e.g. data-control, adaptive depth) @MichaelPoli6

[Stefano Massaroli @Massastrello]

Two of our papers got accepted at #NeurIPS2020!!!
“Dissecting Neural ODEs” (oral)
“Hypersolvers: toward fast continuous depth models” (poster)

[Michael Poli @MichaelPoli6]

As our first NeurIPS experience, I have to say the results surpassed even the wildest of expectations. This is the culmination of a team effort with my dear friend @Massastrello, leading to @Diffeq_ml as an open-source effort for neural differential equations.

[DiffeqML @Diffeq_ml]

A new paper from DiffEqML research group! We speed up Neural ODE inference by learning how to solve them efficiently through *hypersolvers* @Massastrello @MichaelPoli6 The code will be released soon in torchdyn

[Michael Poli @MichaelPoli6]

[1/n] The community has been hard at work to speed up Neural ODEs, e.g. regularization strategies @DavidDuvenaud @chuckberryfinn to keep the ODE easy to solve. We've also been thinking about the same problem, and we propose a different (compatible!) direction. @Massastrello

[DiffeqML @Diffeq_ml]

@DavidDuvenaud @lxuechen Exciting, looks like there is work for us to do

[David Duvenaud @DavidDuvenaud]

We just open-sourced differentiable SDE solvers in PyTorch:
https://t.co/v1f08mjgCq
Now you can put stochastic differential equations in your deep learning models, and neural nets in your SDEs! Credit to @lxuechen.

[DiffeqML @Diffeq_ml]

[1/5] *torchdyn* goes generative! This new release adds continuous normalizing flows @RickyTQChen @wgrathwohl @DavidDuvenaud , energy-based Neural DEs @MilesCranmer @samgreydanus, and extends existing support for higher-order variants. All powered by Lightning! @williamfalcon

[Stefano Massaroli @Massastrello]

We finally got around to open-sourcing more Neural ODE variants in the "torchdyn" library https://t.co/bT6Hm22onx, including our latest "stacked neural ODEs" aka continuous-depth models with piece-wise constant parameters. @MichaelPoli6

[Michael Poli @MichaelPoli6]

[1/4] Excited to share the first experimental release of *torchdyn* https://t.co/BycdsMx9Jf, a PyTorch library for all things neural differential equations! torchdyn is developed by the core DiffEqML team. @Massastrello @Diffeq_ml