Node Duplication Improves Cold-start Link Prediction

2024-02-15

arXiv.org

AI Summary

Graph Neural Networks (GNNs) suffer significant performance degradation in link prediction for low-degree nodes—e.g., cold-start users—limiting their practical deployment in recommender systems. To address this, we propose NodeDup, a lightweight, plug-and-play method that generates self-duplicated replicas for low-degree nodes and establishes intra-replica connections, enabling parameter-free, architecture-agnostic multi-view representation enhancement. NodeDup is the first approach to formalize node self-duplication as a structured multi-view augmentation mechanism specifically for cold-start link prediction, simultaneously improving low-degree node performance while preserving accuracy on high-degree nodes. Extensive experiments across multiple benchmark datasets demonstrate that NodeDup achieves average improvements of 38.49%, 13.34%, and 6.76% in prediction accuracy for isolated, low-degree, and warm-start nodes, respectively—outperforming both state-of-the-art GNNs and dedicated cold-start methods.

Technology Category

• Machine Learning: Multi-instance/Multi-view Learning
• Data Mining & Knowledge Management: Linked Open Data, Knowledge Graphs & KB Completion
• Natural Language Processing: Language Grounding & Multi-modal NLP

Application Category

• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for 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

Graph Neural Networks (GNNs) are prominent in graph machine learning and have shown state-of-the-art performance in Link Prediction (LP) tasks. Nonetheless, recent studies show that GNNs struggle to produce good results on low-degree nodes despite their overall strong performance. In practical applications of LP, like recommendation systems, improving performance on low-degree nodes is critical, as it amounts to tackling the cold-start problem of improving the experiences of users with few observed interactions. In this paper, we investigate improving GNNs' LP performance on low-degree nodes while preserving their performance on high-degree nodes and propose a simple yet surprisingly effective augmentation technique called NodeDup. Specifically, NodeDup duplicates low-degree nodes and creates links between nodes and their own duplicates before following the standard supervised LP training scheme. By leveraging a ''multi-view'' perspective for low-degree nodes, NodeDup shows significant LP performance improvements on low-degree nodes without compromising any performance on high-degree nodes. Additionally, as a plug-and-play augmentation module, NodeDup can be easily applied to existing GNNs with very light computational cost. Extensive experiments show that NodeDup achieves 38.49%, 13.34%, and 6.76% improvements on isolated, low-degree, and warm nodes, respectively, on average across all datasets compared to GNNs and state-of-the-art cold-start methods.

Problem

• Improves GNN performance on low-degree nodes
• Addresses cold-start problem in link prediction
• Enhances user experience with few interactions

Innovation

• Duplicates low-degree nodes for better performance
• Uses multi-view perspective for cold-start improvement
• Plug-and-play module with low computational cost