Advancing Graph Generation through Beta Diffusion

TL;DR

This paper introduces Graph Beta Diffusion, a novel generative model that effectively handles the complex mixed discrete and continuous features of graph data, outperforming existing models on various benchmarks.

Contribution

The paper proposes Graph Beta Diffusion, a new diffusion process tailored for graph data, with a modulation technique to improve graph realism and topology stability.

Findings

Strong performance on multiple graph benchmarks

Effective modeling of discrete and continuous graph features

Enhanced realism and topology stability in generated graphs

Abstract

Diffusion models have excelled in generating natural images and are now being adapted to a variety of data types, including graphs. However, conventional models often rely on Gaussian or categorical diffusion processes, which can struggle to accommodate the mixed discrete and continuous components characteristic of graph data. Graphs typically feature discrete structures and continuous node attributes that often exhibit rich statistical patterns, including sparsity, bounded ranges, skewed distributions, and long-tailed behavior. To address these challenges, we introduce Graph Beta Diffusion (GBD), a generative model specifically designed to handle the diverse nature of graph data. GBD leverages a beta diffusion process, effectively modeling both continuous and discrete elements. Additionally, we propose a modulation technique that enhances the realism of generated graphs by stabilizing critical graph topology while maintaining flexibility for other components. GBD competes strongly with existing models across multiple general and biochemical graph benchmarks, showcasing its ability to capture the intricate balance between discrete and continuous features inherent in real-world graph data. The PyTorch code is available on GitHub.