Permutation Equivariant Neural Controlled Differential Equations for Dynamic Graph Representation Learning
This work addresses dynamic graph modeling for applications like simulated systems and real-world tasks, representing an incremental improvement over existing Graph Neural CDEs.
The paper tackled the problem of dynamic graph representation learning by introducing Permutation Equivariant Neural Graph CDEs, which reduce parameter count while maintaining representational power, leading to improved performance in interpolation and extrapolation tasks.
Dynamic graphs exhibit complex temporal dynamics due to the interplay between evolving node features and changing network structures. Recently, Graph Neural Controlled Differential Equations (Graph Neural CDEs) successfully adapted Neural CDEs from paths on Euclidean domains to paths on graph domains. Building on this foundation, we introduce Permutation Equivariant Neural Graph CDEs, which project Graph Neural CDEs onto permutation equivariant function spaces. This significantly reduces the model's parameter count without compromising representational power, resulting in more efficient training and improved generalisation. We empirically demonstrate the advantages of our approach through experiments on simulated dynamical systems and real-world tasks, showing improved performance in both interpolation and extrapolation scenarios.