Equivariant Diffusion for Molecule Generation in 3D

Emiel Hoogeboom, Vı́ctor Garcia Satorras, Clément Vignac, Max Welling
Proceedings of the 39th International Conference on Machine Learning, PMLR 162:8867-8887, 2022.

Abstract

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.

Cite this Paper

BibTeX
@InProceedings{pmlr-v162-hoogeboom22a, title = {Equivariant Diffusion for Molecule Generation in 3{D}}, author = {Hoogeboom, Emiel and Satorras, V\'{\i}ctor Garcia and Vignac, Cl{\'e}ment and Welling, Max}, booktitle = {Proceedings of the 39th International Conference on Machine Learning}, pages = {8867--8887}, year = {2022}, editor = {Chaudhuri, Kamalika and Jegelka, Stefanie and Song, Le and Szepesvari, Csaba and Niu, Gang and Sabato, Sivan}, volume = {162}, series = {Proceedings of Machine Learning Research}, month = {17--23 Jul}, publisher = {PMLR}, pdf = {https://proceedings.mlr.press/v162/hoogeboom22a/hoogeboom22a.pdf}, url = {https://proceedings.mlr.press/v162/hoogeboom22a.html}, abstract = {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.} }
Endnote
%0 Conference Paper %T Equivariant Diffusion for Molecule Generation in 3D %A Emiel Hoogeboom %A Vı́ctor Garcia Satorras %A Clément Vignac %A Max Welling %B Proceedings of the 39th International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2022 %E Kamalika Chaudhuri %E Stefanie Jegelka %E Le Song %E Csaba Szepesvari %E Gang Niu %E Sivan Sabato %F pmlr-v162-hoogeboom22a %I PMLR %P 8867--8887 %U https://proceedings.mlr.press/v162/hoogeboom22a.html %V 162 %X 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.
APA
Hoogeboom, E., Satorras, V.G., Vignac, C. & Welling, M.. (2022). Equivariant Diffusion for Molecule Generation in 3D. Proceedings of the 39th International Conference on Machine Learning, in Proceedings of Machine Learning Research 162:8867-8887 Available from https://proceedings.mlr.press/v162/hoogeboom22a.html.

Related Material