EXPLORING GRAPH GENERATIVE MODELS TECHNIQUES, APPLICATIONS, AND FUTURE DIRECTIONS

Abstract

Graphs are one of the most elegant ways to store data that shows complex connections and interactions across many entities. Recent progress in deep learning has led to the creation of strong graph-generative models that can learn and create graph-structured data with myriad applications. This article gives an overview of graph-generative models, focusing on the techniques they use, the things they can be used for, and where the field is headed. We talk about well-known methods like Graph Variational Autoencoders (Graph-VAEs), Graph Generative Adversarial Networks (Graph-GANs), and Graph RNN, as well as their variations and enhancements. We also talk about the usefulness of graph-generative models in areas such as drug discovery, studying social networks, finding scams, and studying biological networks. Lastly, we talk about open problems and possible directions for future study in this field that is changing very quickly.