Reproduce “DGR: A General Graph Desmoothing Framework for Recommendation via Global and Local Perspectives” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Leilei Ding, Dazhong Shen, Chao Wang, Tianfu Wang, Le Zhang, Yanyong Zhang
This paper presents DGR, a general graph desmoothing framework designed to mitigate the over-smoothing problem in GCN-based recommendation systems. DGR achieves this by integrating Global Desmoothing Message Passing (GMP) and Local Node Embedding Correction (LEC) modules, which consistently improve the performance of various GCN models across multiple datasets by maintaining distinct user and item embeddings.
Key insights
- Effective desmoothing in GCN-based recommender systems requires a dual approach, addressing both global convergence of embeddings and preserving crucial local collaborative signals.
- The DGR framework is model-agnostic, providing a general, plug-and-play solution that can enhance various existing GCN architectures without requiring extensive redesign.
- By mitigating over-smoothing, DGR allows for the effective use of deeper GCNs, enabling them to capture higher-order collaborative signals without performance degradation.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.