Paper at WWW'21 Workshop

Together with Antonio Carta, Andrea Cossu, and Davide Bacciu, our paper “Catastrophic Forgetting in Deep Graph Networks: an Introductory Benchmark for Graph Classification” has been accepted for publication at the WWW'21 Workshop on Graph Learning Benchmarks. We study whether known continual learning techniques have an impact when applied to graphs, but we also observe that structure-preserving regularization may help. Also, a structure-agnostic baseline shows strong performances in this scenario (again), indicating that there is much to be done at the intersection of the Continual Learning and Graph Representation Learning fields! EDIT: it has also been selected as one of the two orals of the workshop! Great news :)

Federico Errica
Federico Errica
PhD in Computer Science