# Temporal Graph Networks for Deep Learning on Dynamic Graphs

**Authors:** Emanuele Rossi, Ben Chamberlain, Fabrizio Frasca, Davide Eynard, Federico Monti, Michael Bronstein  
**arXiv:** [2006.10637](https://arxiv.org/abs/2006.10637)  
**Date:** 2020-06-18  
**Summary page:** https://www.summarizepaper.com/en/paper/2006.10637v1/

## Key points

- Authors discuss the increasing popularity of Graph Neural Networks (GNNs) in learning complex systems of relations across various domains
- Introduce Temporal Graph Networks (TGNs) as a versatile framework for deep learning on dynamic graphs represented as sequences of timed events
- TGNs leverage a novel combination of memory modules and graph-based operators to outperform existing methods while maintaining computational efficiency
- Detailed ablation study identifies optimal configurations achieving state-of-the-art performance on transductive and inductive prediction tasks for dynamic graphs
- Research offers a comprehensive solution for effectively handling dynamic graph data in applications such as biology, particle physics, social networks, and recommendation systems

## Summary

In their paper titled "Temporal Graph Networks for Deep Learning on Dynamic Graphs," authors Emanuele Rossi, Ben Chamberlain, Fabrizio Frasca, Davide Eynard, Federico Monti, and Michael Bronstein discuss the increasing popularity of Graph Neural Networks (GNNs) in learning complex systems of relations across various domains. The authors introduce Temporal Graph Networks (TGNs), a versatile framework designed for deep learning on dynamic graphs represented as sequences of timed events. TGNs leverage a novel combination of memory modules and graph-based operators to outperform existing methods while maintaining computational efficiency. Through a detailed ablation study, they identify optimal configurations that achieve state-of-the-art performance on transductive and inductive prediction tasks for dynamic graphs. This research contributes significantly to the field by offering a comprehensive solution for effectively handling dynamic graph data in various applications such as biology, particle physics, social networks, and recommendation systems. By introducing TGNs and showcasing their superior performance compared to previous approaches, the authors pave the way for more advanced deep learning techniques tailored to evolving graph structures.