Equivariant Diffusion for Molecule Generation in 3D

Emiel Hoogeboom · Victor Garcia Satorras · ClĂ©ment Vignac · Max Welling

Room 310
[ Abstract ] [ Livestream: Visit Deep Learning: Generative Models/Autoencoders ]
Tue 19 Jul 8:05 a.m. — 8:25 a.m. PDT
[ Paper PDF

This work introduces a diffusion model for molecule generation in 3D that is equivariant to Euclidean transformations. Our E(3) Equivariant Diffusion Model (EDM) learns to denoise a diffusion process with an equivariant network that jointly operates on both continuous (atom coordinates) and categorical features (atom types). In addition, we provide a probabilistic analysis which admits likelihood computation of molecules using our model. Experimentally, the proposed method significantly outperforms previous 3D molecular generative methods regarding the quality of generated samples and the efficiency at training time.

Chat is not available.