A Survey on Graph Diffusion Models: Generative AI in Science for Molecule, Protein and Material

Mengchun Zhang,Maryam Qamar,Taegoo Kang,Yuna Jung,Chenshuang Zhang,Sung-Ho Bae,Chaoning Zhang
DOI: https://doi.org/10.13140/RG.2.2.26493.64480
2023-04-04
Abstract:Diffusion models have become a new SOTA generative modeling method in various fields, for which there are multiple survey works that provide an overall survey. With the number of articles on diffusion models increasing exponentially in the past few years, there is an increasing need for surveys of diffusion models on specific fields. In this work, we are committed to conducting a survey on the graph diffusion models. Even though our focus is to cover the progress of diffusion models in graphs, we first briefly summarize how other generative modeling methods are used for graphs. After that, we introduce the mechanism of diffusion models in various forms, which facilitates the discussion on the graph diffusion models. The applications of graph diffusion models mainly fall into the category of AI-generated content (AIGC) in science, for which we mainly focus on how graph diffusion models are utilized for generating molecules and proteins but also cover other cases, including materials design. Moreover, we discuss the issue of evaluating diffusion models in the graph domain and the existing challenges.
Machine Learning,Computer Vision and Pattern Recognition
What problem does this paper attempt to address?
### Problems the paper attempts to solve This paper aims to review Graph Diffusion Models, especially their applications in molecule, protein and material generation. Specifically, the paper attempts to solve the following problems: 1. **Lack of a specialized review of graph diffusion models**: - Although there are already many review articles on diffusion models, detailed reviews on graph diffusion models are still scarce. Therefore, this paper is committed to filling this knowledge gap. 2. **Unique challenges of graph data**: - Graph data has characteristics such as irregular structure, different scales, diversity and heterogeneity, which make it difficult to directly apply traditional convolution or pooling operations to graph data. In addition, the complexity of graph data requires the development of new methods to effectively process and generate graphs. 3. **Limitations of generative models in graph generation tasks**: - Traditional generative models such as autoregressive models, variational auto - encoders (VAE), normalizing flows and generative adversarial networks (GAN) have certain limitations in graph generation tasks. For example, autoregressive models have difficulty in capturing permutation invariance, VAE has limited scalability on large - scale graphs, and GAN requires likelihood - based optimization, etc. Therefore, it is necessary to explore new methods more suitable for graph generation tasks. 4. **Applications and development of graph diffusion models**: - Graph diffusion models show great potential in fields such as molecule generation, protein generation and material design. The paper hopes to review these models, explore their applications in artificial intelligence - generated content (AIGC) in the scientific field, and discuss the current challenges and future research directions. 5. **Criteria and challenges for evaluating graph diffusion models**: - Evaluating the effectiveness and performance of graph diffusion models is another important issue. Existing evaluation criteria and metrics may not be sufficient to comprehensively evaluate the performance of these models, so it is necessary to explore more appropriate evaluation methods. ### Summary Through the above analysis, the main goal of the paper is to provide researchers with a comprehensive resource to help them better understand and apply graph diffusion models, especially in applications such as molecule, protein and material generation. At the same time, the paper also points out the deficiencies in existing research and proposes future research directions and challenges.