Víctor Garcia (Víctor Garcia Satorras) from UvA Amsterdam his #NeurIPS2019 paper on an hybrid model that combines graphical inference with a learned inverse model. Co-authored with Zeynep Akata & Max Welling . #DLBCN papers.neurips.cc/paper/9532-com…

Wouldn’t it be nice if you could turn any linear transformation into an invertible one? One can take the exponential of a linear transformation, which is guaranteed invertible. Joint work with the brilliant Víctor Garcia Satorras, Jakub Tomczak & Max Welling. Link: arxiv.org/abs/2006.01910

Our paper "The Convolution Exponential and Generalized Sylvester Flows" has been accepted to NeurIPS. Code is now also available at: github.com/ehoogeboom/con…. Joint work with Víctor Garcia Satorras Jakub Tomczak and genius Max Welling.

We are glad to present "E(n) Equivariant Graph Neural Networks". A new simple & effective method to build E(n) equivariance into your graphs. We will present it at ICML. Code is now also available at: github.com/vgsatorras/egnn. Joint work with Emiel Hoogeboom and Max Welling.

Very excited to share our latest work E(n) Equivariant Normalizing Flows for Molecule Generation in 3D. Joint work with Emiel Hoogeboom Fabian Fuchs Ingmar Posner Max Welling. Paper: arxiv.org/abs/2105.09016

Are you interested in E(n) Equivariant Normalizing Flows, which can be used for molecule generation? We wrote a blog that might be interesting ehoogeboom.github.io/post/en_flows/. It covers 1) Normalizing Flows 2) Continuous time flows 3) E(n) GNNs 4) Argmax Flows & 5) E(n) Flows

Virtual conferences are excellent hacks, but they cannot meet the sense of belonging that onsite events provide. Long life to the Deep Learning Barcelona Symposium (Deep Learning Barcelona Symposium). 💪🏻

A short (~50 lines of code) and easy-to-follow implementation of E(n) Equivariant Graph Neural Networks Víctor Garcia Satorras Emiel Hoogeboom Max Welling for HOMO energy prediction. github.com/senya-ashukha/…. Quite a fun weekend project!

Very excited to share this new collaboration! A diffusion model for molecule generation in 3D. We also explore conditional generation in QM9 and evaluate on the larger GEOM-Drugs dataset.

New work on how to construct neural network layers on composite objects of scalars, vectors, bivectors, … --> multivectors! Via Clifford algebras, we generalize convolution and Fourier transforms to multivectors, especially relevant for PDE modeling: arxiv.org/abs/2209.04934

Excited to share our new paper! ✨ Equivariant 3D-Conditional Diffusion Models for Molecular Linker Design arxiv.org/abs/2210.05274 See the thread 🧵

Traveling waves are known to exist throughout the brain in a variety of forms — there are many hypotheses, but their exact computational role is debated. Together with Max Welling we built an RNN which exhibits traveling waves to see what it could do. Here’s what we think: 1/7

Proud to present Denoising Diffusion Models, where we connect the learned score of a diffusion model with force fields to do sampling and simulations🎉 Work done during a wonderful collaborative internship Microsoft Research JCIM & JCTC Journals pubs.acs.org/doi/10.1021/ac…, arxiv.org/abs/2302.00600

Sharing an early preprint of my Microsoft AI4Science summer internship project. We developed SE(3) flow matching for protein backbone generation. Compared to SE(3) diffusion, we find our method achieves higher designability, faster sampling, with a way simpler implementation. 1/8

We have now released the denoising diffusion model codebase! It includes: 🏋️ Pretrained models ✅ Easy sampling and evaluation 💻 Full model and training scripts Check out our repository here: github.com/microsoft/two-…

If you trained on the true equilibrium distribution structures extracted from a simulation you get the true forces (if your noise is sufficiently low) This paper first showed this and we have validated for a simple system that it is precise. arxiv.org/abs/2302.00600

The afternoon session on the Deep Generative Models days continues with Diffusion Models with Chin-Wei Huang & Víctor Garcia Satorras from Microsoft Research. Let’s learn what is new around Diffusion Models and how they are used today #ProbAI

Two very interesting papers related to few-shot learning on arXiv today: - arxiv.org/abs/1711.04340 - arxiv.org/abs/1711.04043 Unfortunately harmfully close to my current work... so I'm also having a rough morning :-)

Did you know that there is an engineering school in Barcelona where THREE BSc/MSc theses have been accepted in #iclr2018 ? A milestone for @TelecomBCN Universitat Politècnica de Catalunya (UPC), in partnership with NYC Columbia | EE & NYU Data Science. Congrats Víctor Campos, Víctor Garcia & Àlex Nowak !

Deep interest on the few-shot learning with graph neural networks by Víctor Garcia, former student in our lab Universitat Politècnica de Catalunya (UPC) Barcelona #DLUPC #ICLR2018