GRAPH AUTOENCODERS: A SURVEY

Graph analysis serves as a robust approach for the in-depth exploration of the inherent characteristics of graph data. Nonetheless, due to the non-Euclidean nature of such data, conventional data analysis techniques often incur significant computational expenses and spatial overhead. Graph autoencoders present a viable solution to the challenges associated with graph analysis by converting the original graph data into a low-dimensional representation while maintaining essential information. This transformation subsequently improves the efficacy of various downstream tasks, including node classification, link prediction, and node clustering. This paper offers a thorough review of the existing literature on graph autoencoders, encapsulating the fundamental strategies employed by these models and their applications in downstream tasks. Additionally, the paper suggests prospective avenues for future research in the domain of graph autoencoders.