Dynamic Graph Recommendation via Sparse Augmentation and Singular Adaptation

2025-11-14

AI Summary

In dynamic recommendation, large-scale graph fine-tuning faces dual challenges: prohibitive computational overhead and inadequate representation of sparse nodes—both stemming from degree distribution long-tails induced by interaction sparsity. To address this, we propose GraphSASA, a parameter-efficient dynamic graph recommendation framework. GraphSASA enhances robustness of sparse-node representations via test-time hierarchical graph augmentation and introduces singular-value adaptive fine-tuning: it freezes the backbone parameters and optimizes only a low-rank singular-value subspace. This strategy reduces trainable parameters by over 90% while significantly improving long-tail node modeling. Evaluated on three large-scale dynamic graph datasets, GraphSASA achieves state-of-the-art recommendation performance at substantially lower computational cost, effectively balancing accuracy and efficiency.

Technology Category

• Machine Learning: Graph-based Machine Learning
• Search and Optimization: Learning to Search
• Data Mining & Knowledge Management: Graph Mining, Social Network Analysis & Community

Application Category

• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs
• User Modeling, Personalization and Recommendation: Federated recommendation systems and personalization
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking

Abstract

Dynamic recommendation, focusing on modeling user preference from historical interactions and providing recommendations on current time, plays a key role in many personalized services. Recent works show that pre-trained dynamic graph neural networks (GNNs) can achieve excellent performance. However, existing methods by fine-tuning node representations at large scales demand significant computational resources. Additionally, the long-tail distribution of degrees leads to insufficient representations for nodes with sparse interactions, posing challenges for efficient fine-tuning. To address these issues, we introduce GraphSASA, a novel method for efficient fine-tuning in dynamic recommendation systems. GraphSASA employs test-time augmentation by leveraging the similarity of node representation distributions during hierarchical graph aggregation, which enhances node representations. Then it applies singular value decomposition, freezing the original vector matrix while focusing fine-tuning on the derived singular value matrices, which reduces the parameter burden of fine-tuning and improves the fine-tuning adaptability. Experimental results demonstrate that our method achieves state-of-the-art performance on three large-scale datasets.

Problem

• Addresses computational inefficiency in dynamic graph recommendation fine-tuning
• Solves insufficient representation for sparse interaction nodes in graphs
• Reduces parameter burden while maintaining fine-tuning adaptability

Innovation

• Test-time augmentation enhances node representation distributions
• Singular value decomposition reduces fine-tuning parameter burden
• Freezing original vector matrix improves fine-tuning adaptability