arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2502.08071cs.IR
收藏

协同过滤遇见谱移:连接用户-物品交互与图结构侧信息

Collaborative Filtering Meets Spectrum Shift: Connecting User-Item Interaction with Graph-Structured Side Information

Yunhang He, Cong Xu, Jun Wang, Wei Zhang

AI总结:

针对图结构侧信息整合到用户-物品二部图后现有图协同过滤方法性能下降的谱移问题,提出谱移校正方法SSC,无需定制设计即可适配任意图结构侧信息,在推荐任务中实现最高23%的相对提升且无额外计算开销。

摘要:

图神经网络(GNNs)已在协同过滤领域展现出优越性,其中用户-物品(U-I)交互二部图是基础数据格式。然而,当图结构侧信息(例如多模态相似度图或社交网络)被整合到U-I二部图中时,现有图协同过滤方法无法取得令人满意的性能。我们从谱的角度对该问题进行了定量分析。回顾可知,二部图在[-1, 1]范围内拥有完整频谱,最高频率恰好可在-1处达到,最低频率在1处;但我们观察到,随着更多侧信息被纳入,增广邻接矩阵的最高频率会逐渐右移。这种谱移现象导致此前为[-1, 1]完整频谱构建的方法会为不同频率分配不匹配的重要性。为此,我们提出了谱移校正(Spectrum Shift Correction,简称SSC)方法,通过引入移位和缩放因子使谱GNN能够适应发生偏移的频谱。与以往需要针对不同数据类型进行定制化设计的侧信息利用范式不同,SSC直接将传统图协同过滤与任意图结构侧信息连接起来。在社交推荐和多模态推荐任务上的实验证明了SSC的有效性,其在不产生任何额外计算开销的情况下取得了最高达23%的相对提升。我们的代码可在https://github.com/yhhe2004/SSC-KDD获取。

英文摘要:

Graph Neural Networks (GNNs) have demonstrated their superiority in collaborative filtering, where the user-item (U-I) interaction bipartite graph serves as the fundamental data format. However, when graph-structured side information (e.g., multimodal similarity graphs or social networks) is integrated into the U-I bipartite graph, existing graph collaborative filtering methods fall short of achieving satisfactory performance. We quantitatively analyze this problem from a spectral perspective. Recall that a bipartite graph possesses a full spectrum within the range of [-1, 1], with the highest frequency exactly achievable at -1 and the lowest frequency at 1; however, we observe as more side information is incorporated, the highest frequency of the augmented adjacency matrix progressively shifts rightward. This spectrum shift phenomenon has caused previous approaches built for the full spectrum [-1, 1] to assign mismatched importance to different frequencies. To this end, we propose Spectrum Shift Correction (dubbed SSC), incorporating shifting and scaling factors to enable spectral GNNs to adapt to the shifted spectrum. Unlike previous paradigms of leveraging side information, which necessitate tailored designs for diverse data types, SSC directly connects traditional graph collaborative filtering with any graph-structured side information. Experiments on social and multimodal recommendation demonstrate the effectiveness of SSC, achieving relative improvements of up to 23% without incurring any additional computational overhead. Our code is available at https://github.com/yhhe2004/SSC-KDD.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑