1 Title

        Self-Supervised Hypergraph Transformer for Recommender Systems（Lianghao Xia, Chao Huang, Chuxu Zhang）【KDD 2022】

2 Conclusion

        User behavior data in many practical recommendation scenarios is often noisy and exhibits skewed distribution, which may result in suboptimal representation performance in GNN-based
models. In this paper, we propose SHT, a novel Self-Supervised Hypergraph Transformer framework (SHT) which augments user representations by exploring the global collaborative relationships in an explicit way. Specifically, we first empower the graph neural CF paradigm to maintain global collaborative effects among users and items with a hypergraph transformer network. With the distilled global context, a cross-view generative self-supervised learning component is proposed for data augmentation over the user-item interaction graph, so as to enhance the robustness of recommender systems.

3 Good Sentences

        1、Despite the effectiveness of the above graph-based CF models by providing state-of-the-art recommendation performance, several key challenges have not been well addressed in existing methods. First, data noise is ubiquitous in many recommendation scenarios due to a variety of factors. For example, users may click their uninterested products due to the over-recommendation of popular items（The shortcomings of previous works ）
        2、produces state-of-the-art performance by generating contrastive views with randomly node and edge dropout operations. Following this research line, HCCF leverages the hypergraph to generate contrastive signals to improve the graph-based recommender system. Different from them, this work enhances the graph-based collaborative filtering paradigm with a generative selfsupervised learning framework.（The improvement and creativity of this study has done）
        3、The variant without self-augmented learning yields obvious performance degradation in all cases, which validates the positive effect of our augmented global-to-local knowledge transferring. The effect of our meta-network-based domain adaption can also be observed in the variant -Meta.（The usage of Self-Supervised Learning）

---

SHT共由四部分组成：

1. 用户-物品的交互图；
2. 节点通过两次聚合邻居节点信息，生成节点embedding表示；
3. 节点embedding与超边embedding之间相互聚合，经过迭代，不断优化各自embedding表示；超边embedding由节点embedding生成；
4. 利用超边与节点信息进行数据增强