Abstract

Graph Convolution Network (GCN) has become new state-of-the-art for collaborative filtering. Nevertheless, the reasons of its effectiveness for recommendation are not well understood. Existing work that adapts GCN to recommendation lacks thorough ablation analyses on GCN, which is originally designed for graph classification tasks and equipped with many neural network operations. However, we empirically find that the two most common designs in GCNs -- feature transformation and nonlinear activation -- contribute little to the performance of collaborative filtering. Even worse, including them adds to the difficulty of training and degrades recommendation performance.

Keywords

Computer scienceCollaborative filteringConvolution (computer science)GraphRecommender systemArtificial intelligenceArtificial neural networkMachine learningTheoretical computer science

Affiliated Institutions

Related Publications

Fuel Cell Systems Explained

Andrew Dicks , D.A.J. Rand

Preface. Foreword to the first edition. Acknowledgements. Abbreviations. Symbols. Introduction. Efficiency and Open Circuit Voltage. Operational Fuel Cell Voltages. Proton Excha...

2018 4195 citations

Publication Info

Year
2020
Type
article
Pages
639-648
Citations
3523
Access
Closed

External Links

View on DOI.org

Social Impact

Altmetric
PlumX Metrics

Social media, news, blog, policy document mentions

Citation Metrics

3523
OpenAlex

Cite This

Xiangnan He, Kuan Deng, Xiang Wang et al. (2020). LightGCN. , 639-648. https://doi.org/10.1145/3397271.3401063

Identifiers

DOI
10.1145/3397271.3401063