A Systematic Survey on Deep Generative Models for Graph Generation

Abstract

Graphs are important data representations for describing objects and their relationships, which appear in a wide diversity of real-world scenarios. As one of a critical problem in this area, graph generation considers learning the distributions of given graphs and generating more novel graphs. Owing to its wide range of applications, generative models for graphs have a rich history, which, however, are traditionally hand-crafted and only capable of modeling a few statistical properties of graphs. Recent advances in deep generative models for graph generation is an important step towards improving the fidelity of generated graphs and paves the way for new kinds of applications. This article provides an extensive overview of the literature in the field of deep generative models for graph generation. Firstly, the formal definition of deep generative models for the graph generation as well as preliminary knowledge is provided. Secondly, two taxonomies of deep generative models for unconditional, and conditional graph generation respectively are proposed; the existing works of each are compared and analyzed. After that, an overview of the evaluation metrics in this specific domain is provided. Finally, the applications that deep graph generation enables are summarized and five promising future research directions are highlighted.

Discussion of “Specifying prior distributions in reliability applications”: Towards new formal rules for informative prior elicitation?

“…This does not forbid prior and posterior sampling (e.g., by approximation and numerical methods 54,55 ). Conceptually, the approach which consists in defining $$$ \pi \left(\theta \right) $$$ as the limit distribution of a simulation algorithm is close to generative tools currently used in machine learning (e.g., Reference 56). However, obtaining explicit forms for the prior $$$ \pi \left(\theta |m,\omega (m)\right) $$$ can be useful to conduct several important tasks with limited computational effort, as calibration tasks, sensitivity analyses, and so forth.…”

GEMS: Scene Expansion using Generative Models of Graphs

“…Generative Models of Graphs Graphs are a powerful and natural representation of the data in many application settings. And, as with many other domains, generative models trained over a set of observed graphs have received much recent attention [16]. Most existing work considers molecular graphs, where sampling from a trained model allows the generation of novel molecules, the core objective of drug design.…”

MARS: Markov Molecular Sampling for Multi-objective Drug Discovery

“…Recent years have witnessed the success of applying deep generative models and molecular graph representation learning in drug discovery (Schwalbe-Koda & Gómez-Bombarelli, 2020;Guo & Zhao, 2020). Existing approaches for molecular property optimization can be grouped into four categories, including generation with a) Bayesian inference, b) reinforcement learning, c) encoderdecoder translation models, and d) evolutionary and genetic algorithms.…”