TCS Seminar - An Introduction to Temporal Graph Neural Networks

This talk introduces Temporal Graph Neural Networks (TGNNs) as models for learning from systems that combine relational structure and temporal dynamics. Starting from basic neural network concepts, we build up to graph neural networks, message passing, and temporal graph settings, distinguishing between static graphs with time-varying features and fully evolving graph structures. Through examples in synthetic forecasting, traffic prediction, and brain activity modeling, the tutorial illustrates when graph information can improve temporal learning and how TGNNs capture spatio-temporal dependencies in real-world networks.