From static to dynamic Graph Neural Networks

In recent years, Graph Neural Networks (GNNs) have emerged as powerful tools for analyzing and learning from graph-structured data. Static GNNs have proven remarkably effective in tasks like node classification and link prediction. In the last few years, these architectures were extended to the dynamic graph setting, a good representation of many real-world systems that evolve over time.

In 2021, I was a collaborator in the Pytorch Geometric Temporal (PGT) project [1], which is a deep learning framework combining state-of-the-art machine learning algorithms for neural spatiotemporal signal processing. The main goal of the PGT library is to make temporal geometric deep learning available for researchers and machine learning practitioners in a unified, easy-to-use framework.

In this presentation, I aim to give a partial overview of GNNs, focusing on the best-performing algorithms implemented in the PGT library.

[1] Rozemberczki, Benedek, et al. "Pytorch geometric temporal: Spatiotemporal signal processing with neural machine learning models." Proceedings of the 30th ACM International Conference on Information & Knowledge Management. 2021.