Generative Diffusion Models on Graphs: Methods and Applications

• DOI: 10.48550/arxiv.2302.02591
• Publication Date: 2023-01-01
• Authors: Wenqi Fan; Cheng-Yi Liu; Yunqing Liu; Jiatong Li; Hang Li; Hui Liu; Jiliang Tang; Qing Li
• Keywords: Generative model; Leverage (statistics); Inpainting

Diffusion models, as a novel generative paradigm, have achieved remarkable success in various image generation tasks such inpainting, image-to-text translation, and video generation. Graph is crucial computational task on graphs with numerous real-world applications. It aims to learn the distribution of given then generate new graphs. Given great diffusion models generation, increasing efforts been made leverage these techniques advance graph recent years. In this paper, we first provide comprehensive overview graphs, particular, review representative algorithms for three variants i.e., Score Matching Langevin Dynamics (SMLD), Denoising Probabilistic Model (DDPM), Score-based Generative (SGM). Then, summarize major applications specific focus molecule protein modeling. Finally, discuss promising directions graph-structured data. For survey, also created GitHub project website by collecting supporting resources at link: https://github.com/ChengyiLIU-cs/Generative-Diffusion-Models-on-Graphs