'Si'multaneous 'S'patial-'T'emporal Message Passing for Dynamic Graph Representation Learning

2026-05-25

AI Summary

This work proposes SiST-GNN, a novel dynamic graph neural network that integrates spatial and temporal information within a single message-passing step, overcoming the limitations of conventional approaches that process these dimensions sequentially. By maintaining recursive hidden states for each node and pairing them with current features to construct cross-temporal edges, SiST-GNN performs graph convolution synchronously on an augmented graph, enabling direct incorporation of historical neighbor trajectories into current representation learning. This unified spatiotemporal architecture eliminates the need for separate temporal and spatial modules. Empirical results demonstrate substantial gains: on 14 datasets for link prediction, SiST-GNN outperforms the strongest baselines by 109–277% under fixed splits and by 68–194% in real-time updating scenarios; in dynamic node classification, it surpasses discrete-time methods by 7–22% and achieves performance comparable to continuous-time models.

Abstract

Dynamic graph neural networks (DGNNs) that operate on snapshot sequences typically fall into one of two categories. \emph{Temporal-first} approaches build per-node temporal embeddings and only afterwards perform spatial aggregation, whereas \emph{Spatial-first} approaches invert this order, feeding the output of a graph convolution into a downstream temporal module. In either case, the rigid sequencing forces the second stage to consume an already-compressed summary produced by the first, ruling out joint reasoning over topology and evolution; concretely, the message-passing operator never gets to weight a neighbor's contribution by that neighbor's \emph{past} trajectory. This paper introduces \textbf{SiST-GNN} (\textbf{Si}multaneous \textbf{S}patial-\textbf{T}emporal \textbf{GNN}), which fuses the two signals inside a single message-passing operation rather than chaining them. Concretely, at each snapshot we maintain a recurrent hidden state per node that summarises its history, pair it with the node's current feature vector, and treat the pair as two nodes joined by a cross-time edge; running a standard graph convolution on this temporally augmented graph yields the updated representation. Our empirical study spans nine public baselines and fourteen model-dataset combinations, covering both fixed-split and live-update evaluation regimes. Across every public benchmark, SiST-GNN sets a new state of the art in link prediction task over the strongest prior method by $109$--$277\%$ in the fixed-split setting and by $68$--$194\%$ in the live-update setting. We additionally construct three dynamic node-classification tasks by discretising the underlying continuous-time event streams; here SiST-GNN beats the leading discrete-time (DTDG) baseline by $7$--$22\%$ and matches continuous-time (CTDG) methods that consume the raw events directly.

Problem

• Dynamic Graph Neural Networks
• Spatial-Temporal Message Passing
• Joint Reasoning
• Temporal Trajectory
• Graph Representation Learning

Innovation

• Simultaneous Spatial-Temporal
• Dynamic Graph Neural Networks
• Message Passing
• Temporal Augmentation
• Joint Reasoning