Heterogeneous Graph Masked Contrastive Learning for Robust Recommendation

2025-05-30

AI Summary

In heterogeneous graph recommendation, existing methods suffer from dense structural noise in meta-path construction and noise amplification during heterogeneous graph neural network (HGNN) propagation, leading to non-robust node embeddings. To address this, we propose Masked Contrastive Learning (MCL), the first framework to incorporate random edge masking into heterogeneous graph contrastive learning. MCL constructs dual views—single-hop neighbors and meta-path-based neighbors—and jointly models local and high-order semantics under graph structure augmentation. It requires no auxiliary supervision, simultaneously optimizing structural denoising and semantic consistency. Extensive experiments on three real-world datasets demonstrate that MCL consistently outperforms state-of-the-art HGNN baselines, achieving average improvements of 3.2–5.8% in AUC and Recall. Ablation studies and robustness analysis further validate its superior noise resilience and generalization stability.

Technology Category

• Machine Learning: Graph-based Machine Learning
• Data Mining & Knowledge Management: Graph Mining, Social Network Analysis & Community
• Humans and AI: Human-in-the-loop Machine Learning

Application Category

• Graph Algorithms and Modeling for the Web: Foundation models and LLMs 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

Heterogeneous graph neural networks (HGNNs) have demonstrated their superiority in exploiting auxiliary information for recommendation tasks. However, graphs constructed using meta-paths in HGNNs are usually too dense and contain a large number of noise edges. The propagation mechanism of HGNNs propagates even small amounts of noise in a graph to distant neighboring nodes, thereby affecting numerous node embeddings. To address this limitation, we introduce a novel model, named Masked Contrastive Learning (MCL), to enhance recommendation robustness to noise. MCL employs a random masking strategy to augment the graph via meta-paths, reducing node sensitivity to specific neighbors and bolstering embedding robustness. Furthermore, MCL employs contrastive cross-view on a Heterogeneous Information Network (HIN) from two perspectives: one-hop neighbors and meta-path neighbors. This approach acquires embeddings capturing both local and high-order structures simultaneously for recommendation. Empirical evaluations on three real-world datasets confirm the superiority of our approach over existing recommendation methods.

Problem

• HGNNs propagate noise in dense graphs affecting recommendations
• MCL enhances robustness via meta-path masking and contrastive learning
• MCL captures local and high-order structures for better embeddings

Innovation

• Random masking strategy for graph augmentation
• Contrastive cross-view on Heterogeneous Information Network
• Captures local and high-order structures simultaneously