Jean Feydy (@FeydyJean) profile picture

Researcher at INRIA Paris: geometric (deep) learning, optimal transport and computational anatomy.

Paris, France
Joined July 2019
Photo by Jean Feydy (@FeydyJean)Photo by Jean Feydy (@FeydyJean)Photo by Jean Feydy (@FeydyJean)Photo by Jean Feydy (@FeydyJean)Photo by Jean Feydy (@FeydyJean)Photo by Jean Feydy (@FeydyJean)Photo by Jean Feydy (@FeydyJean)Photo by Jean Feydy (@FeydyJean)Photo by Jean Feydy (@FeydyJean)Photo by Jean Feydy (@FeydyJean)
KeOps 1.1 is out! We introduce an exact `LazyTensor` decorator for @PyTorch and #NumPy: free x30 speedup + linear memory footprint for kernel methods, mesh processing and Sinkhorn on the GPU. kernel-operations.io
3
53
152
Fast end-to-end learning on protein surfaces (w. @FreyrSverrisson, @befcorreia and @mmbronstein) will be live soon at #CVPR2021. Join us for a chat at 10pm-12.30am EDT (Paper Session 11). openaccess.thecvf.com/conten…
3
16
2
64
Key ingredient 1: a fast sampling layer for protein surfaces.
1
5
Key ingredient 2: fast quasi-geodesic convolutions on oriented point clouds, implemented using KeOps symbolic matrices. kernel-operations.io
6
A most original use of GeomLoss on 3D shape textures, with pairwise Wasserstein distances between 1,000 curvature histograms. Enabling the use of OT as a "standard" tool for large-scale data analysis is a major target for 2021! @RFlamary
Replying to @AnnaSongMaths
To visualize the space of 3D shape textures, we combine UMAP with the optimal transport distance on 2D curvature diagrams. @leland_mcinnes @gabrielpeyre
6
25
Membranes and tubules are part of the same family! These 3D shape textures optimize curvature functionals that generalize the traditional Willmore and Helfrich energies. arxiv.org/abs/2103.04856
2
8
3
28
our paper with @befcorreia @FreyrSverrisson @FeydyJean on fast learning on proteins is accepted to @CVPR End-to-end differentiable architecture, precomputations on the fly, more accurate and x100 faster than #MaSIF biorxiv.org/content/10.1101/… Code coming soon
2
29
155
New #MaSIF architecture for #geometricdeeplearning on proteins with @befcorreia @FreyrSverrisson @FeydyJean works directly on the atomic point cloud, computes the molecular surface on the fly, is end-to-end differentiable, and runs 10x-100x faster biorxiv.org/content/10.1101/…
1
41
1
179
KeOps symbolic tensors now accelerate (x10-x100) a wide range of methods, from UMAP embeddings to geometric neural networks. B. Charlier, @JoanGlaunes, @mmbronstein and myself will be happy to discuss it this Wednesday at #NeurIPS2020: neurips.cc/virtual/2020/publ…
1
26
3
137
Optimal Transport generalizes sorting to spaces of dimension D > 1. GeomLoss implements a smooth Quicksort-like solver, packaged as a PyTorch layer. Keeps a linear memory usage + x10 to x1,000 speedup compared with the Auction and Sinkhorn algorithms. https://xitter.cf/t.co/HhqtuDX3TX
4
96
7
458
The video above was generated on a free Colab notebook with clouds of ~200k points: solving the required 3*10 regularized OT problems took 40s. The low-res animation below - 3*4 regularized OT problems betweens clouds of 10k points - was computed in <1s. https://xitter.cf/t.co/ltnlXbk6Sy
1
4
25
KeOps now fully supports automatic differentiation, batch processing and LogSumExp or K-Min reductions. Check our tutorials on spectral clustering, optimal transport and others! Joint work with Benjamin Charlier and @JoanGlaunes.
1
6
18