Trajectory Encoding Temporal Graph Networks

2025-04-15

AI Summary

Existing temporal graph networks (TGNs) face a fundamental trade-off between transductive and inductive learning: anonymous models exhibit strong generalization but cannot distinguish known nodes, whereas non-anonymous models achieve high transductive accuracy yet struggle to adapt to unseen nodes. To address this, we propose TrajEncoder—a novel trajectory encoding mechanism that introduces automatically extensible temporal positional identifiers as dynamic node IDs, enabling modeling of historical behavioral paths without requiring fixed node features. TrajEncoder integrates timestamp-aware message passing, multi-head attention-based fusion, and learnable ID embeddings, thereby unifying the advantages of both anonymous and non-anonymous TGNs for the first time. Evaluated on three real-world dynamic graph datasets, our method consistently outperforms state-of-the-art approaches on link prediction and node classification tasks. It achieves up to 12.7% improvement in inductive performance while simultaneously attaining significant gains in transductive accuracy.

Technology Category

• Machine Learning: Graph-based Machine Learning
• Planning, Routing, and Scheduling: Temporal Planning
• Knowledge Representation and Reasoning: Geometric, Spatial, and Temporal Reasoning

Application Category

• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking

Abstract

Temporal Graph Networks (TGNs) have demonstrated significant success in dynamic graph tasks such as link prediction and node classification. Both tasks comprise transductive settings, where the model predicts links among known nodes, and in inductive settings, where it generalises learned patterns to previously unseen nodes. Existing TGN designs face a dilemma under these dual scenarios. Anonymous TGNs, which rely solely on temporal and structural information, offer strong inductive generalisation but struggle to distinguish known nodes. In contrast, non-anonymous TGNs leverage node features to excel in transductive tasks yet fail to adapt to new nodes. To address this challenge, we propose Trajectory Encoding TGN (TETGN). Our approach introduces automatically expandable node identifiers (IDs) as learnable temporal positional features and performs message passing over these IDs to capture each node's historical context. By integrating this trajectory-aware module with a standard TGN using multi-head attention, TETGN effectively balances transductive accuracy with inductive generalisation. Experimental results on three real-world datasets show that TETGN significantly outperforms strong baselines on both link prediction and node classification tasks, demonstrating its ability to unify the advantages of anonymous and non-anonymous models for dynamic graph learning.

Problem

• Balancing transductive and inductive performance in dynamic graphs
• Overcoming limitations of anonymous and non-anonymous TGN designs
• Unifying node distinction and generalization in temporal graph networks

Innovation

• Automatically expandable node IDs as features
• Message passing over IDs for historical context
• Multi-head attention integrates trajectory-aware module