A Transfer Framework for Enhancing Temporal Graph Learning in Data-Scarce Settings

2025-03-02

AI Summary

Temporal Graph Neural Networks (TGNNs) suffer from performance degradation under scarce training data, limiting their practical deployment. Method: This paper proposes a transferable, structured bipartite encoding framework tailored for TGNNs, which—uniquely—decouples node memory from feature representation. It introduces a reusable memory module and inductive patterns across datasets, enabling cross-domain knowledge transfer. The framework leverages unsupervised pretraining followed by low-resource fine-tuning to facilitate efficient knowledge transfer across heterogeneous temporal graphs. Contribution/Results: On real-world benchmarks under low-data regimes, our approach achieves a 56% improvement over non-transfer baselines and outperforms existing transfer methods by 36%. It effectively addresses the critical limitation of poor cross-dataset generalizability in TGNNs, significantly advancing their applicability in data-scarce scenarios.

Technology Category

• Machine Learning: Graph-based Machine Learning
• Data Mining & Knowledge Management: Mining of Spatial, Temporal or Spatio-Temporal Data
• Knowledge Representation and Reasoning: Geometric, Spatial, and Temporal Reasoning

Application Category

• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs
• Semantics and Knowledge: Scalable techniques for the creation, curation, publication, maintenance, and consumption of large, Web-based, structured, reusable, knowledge graphs and ontologies
• Web Mining and Content Analysis: Bridging structured and unstructured data

Abstract

Dynamic interactions between entities are prevalent in domains like social platforms, financial systems, healthcare, and e-commerce. These interactions can be effectively represented as time-evolving graphs, where predicting future connections is a key task in applications such as recommendation systems. Temporal Graph Neural Networks (TGNNs) have achieved strong results for such predictive tasks but typically require extensive training data, which is often limited in real-world scenarios. One approach to mitigating data scarcity is leveraging pre-trained models from related datasets. However, direct knowledge transfer between TGNNs is challenging due to their reliance on node-specific memory structures, making them inherently difficult to adapt across datasets. To address this, we introduce a novel transfer approach that disentangles node representations from their associated features through a structured bipartite encoding mechanism. This decoupling enables more effective transfer of memory components and other learned inductive patterns from one dataset to another. Empirical evaluations on real-world benchmarks demonstrate that our method significantly enhances TGNN performance in low-data regimes, outperforming non-transfer baselines by up to 56% and surpassing existing transfer strategies by 36%

Problem

• Enhancing temporal graph learning in data-scarce settings.
• Overcoming challenges in transferring knowledge between temporal graph neural networks.
• Improving TGNN performance with a novel transfer approach.

Innovation

• Structured bipartite encoding mechanism for node representation
• Decoupling node features for effective knowledge transfer
• Enhanced TGNN performance in low-data scenarios