nuist_Jiang的博客

from Xiangnan He ,University of Science and Technology of China,published in SIGIR‘20， July 25-30,Virtual Event, China.研究背景GCN（图卷积网络）成为协同过滤的最新技术，但是推荐的有效性原因不明确，现有工作缺乏对GCN的彻底的消融分析。GCN中最常见的两个设计-特征转换和非线性激活-对协作过滤的性能几乎没有贡献。更糟糕的是，包括它们会增加培训的难度并降低推荐性能。GCN具体是做什

Neural Graph Collaborative Filteringauthor: Xiang Wang, Xiangnan He, Meng Wangfrom： National University of Singaporepublished in SIGIR’19, July 21-25, 2019, Paris, FranceResearch Background:Learning vector representations (aka. embeddings) of users an