DiGress: Discrete Denoising diffusion for graph generation

TL;DR

DiGress introduces a novel discrete diffusion model for graph generation that improves sample quality and scalability, achieving state-of-the-art results on molecular datasets and enabling large-scale drug molecule generation.

Contribution

It presents the first scalable discrete diffusion model for graph generation that incorporates a Markovian noise process and auxiliary features, advancing graph generative modeling.

Findings

Achieves up to 3x validity improvement on planar graphs.

Scales to 1.3 million drug-like molecules without specialized representations.

Sets new state-of-the-art performance on molecular graph datasets.

Abstract

This work introduces DiGress, a discrete denoising diffusion model for generating graphs with categorical node and edge attributes. Our model utilizes a discrete diffusion process that progressively edits graphs with noise, through the process of adding or removing edges and changing the categories. A graph transformer network is trained to revert this process, simplifying the problem of distribution learning over graphs into a sequence of node and edge classification tasks. We further improve sample quality by introducing a Markovian noise model that preserves the marginal distribution of node and edge types during diffusion, and by incorporating auxiliary graph-theoretic features. A procedure for conditioning the generation on graph-level features is also proposed. DiGress achieves state-of-the-art performance on molecular and non-molecular datasets, with up to 3x validity improvement on a planar graph dataset. It is also the first model to scale to the large GuacaMol dataset containing 1.3M drug-like molecules without the use of molecule-specific representations.