<ul><li>Dynamic graphs provide a flexible data abstraction for modelling real-world systems.</li><li>Graph neural networks (GNNs) are powerful tools for prediction and inference on dynamic graphs.</li><li>This work proposes using a dynamic graph representation called the unfolding for valid prediction sets via conformal prediction.</li><li>The approach achieves valid prediction sets with minimal assumptions and provides real data examples demonstrating improved accuracy.</li></ul>

Valid Conformal Prediction for Dynamic GNNs

Discover more