Reproduce “Enhancing Recommendation Systems with GNNs and Addressing Over-Smoothing” 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

Wenyi Liu, Ziqi Zhang, Xinshi Li, Jiacheng Hu, Yuanshuai Luo, Junliang Du

Researchers developed a Graph Neural Network (GNN) based recommendation algorithm incorporating initial residual connections and identity mapping to mitigate over-smoothing and enhance information flow. This approach achieved superior Recall@20 and NDCG@20 scores compared to other GNN and traditional baselines across Gowalla, Yelp-2018, and Amazon-Book datasets.

Key insights

  • Initial residual connections effectively combat over-smoothing by preserving initial, distinct node information, enabling deeper GNNs to capture higher-order collaborative signals.
  • Identity mapping further stabilizes information flow in deep GNNs, allowing non-linear learning without hindering feature distinctiveness and preventing information loss.
  • Aggregating embeddings from all GNN layers, from shallow to deep, allows the model to leverage a comprehensive range of collaborative signals, combining both local and global information.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.