Hypergraph Contrastive Collaborative Filtering

Collaborative Filtering (CF) has emerged as fundamental paradigms for users and items into latent representation space, with their patterns from interaction data. Among various CF techniques, the of GNN-based recommender systems, e.g., PinSage and LightGCN, has the state-of-the-art performance. However, two key challenges have not well explored in existing solutions: i) The over-smoothing effect with graph-based CF architecture, may cause the indistinguishable user and degradation of recommendation results. ii) The supervision (i.e., user-item interactions) are usually scarce and skewed in reality, which limits the representation power of CF paradigms. tackle these challenges, we propose a new self-supervised recommendation Hypergraph Contrastive Collaborative Filtering (HCCF) to jointly local and global collaborative relations with a hypergraph-enhanced-view contrastive learning architecture. In particular, the designed structure learning enhances the discrimination ability of GNN-based paradigm, so as to comprehensively capture the complex high-order among users. Additionally, our HCCF model effectively integrates hypergraph structure encoding with self-supervised learning to reinforce representation quality of recommender systems, based on the-enhanced self-discrimination. Extensive experiments on three datasets demonstrate the superiority of our model over various-of-the-art recommendation methods, and the robustness against sparse user data. Our model implementation codes are available at://github.com/akaxlh/HCCF.