Advancing Graph Generation through Beta Diffusion
This work addresses a domain-specific problem for researchers and practitioners in graph generation, offering an incremental improvement over existing diffusion models by better handling the statistical patterns of graph data.
The paper tackles the challenge of generating graphs with mixed discrete and continuous components by introducing Graph Beta Diffusion (GBD), which uses a beta diffusion process and modulation technique to improve realism and stability, achieving strong performance on general and biochemical graph benchmarks.
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.