Reproduce “AlignGroup: Learning and Aligning Group Consensus with Member Preferences for Group Recommendation” 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
Jinfeng Xu, Zheyu Chen, Jinze Li, Shuo Yang, Hewei Wang, Edith C.-H. Ngai
AlignGroup introduces a hypergraph neural network with a self-supervised alignment task to jointly model group consensus and individual member preferences for group recommendation. This approach yields up to 87.70% higher NDCG@5 on the CAMRa2011 dataset and 14.63% higher NDCG@5 on Mafengwo compared to state-of-the-art baselines, demonstrating superior fine-grained recommendation accuracy.
Key insights
- Explicitly aligning learned group consensus with common member preferences via a self-supervised task significantly enhances the model's ability to provide fine-grained and accurately ranked recommendations.
- Modeling inter-group relationships, such as similarity based on member overlap, is crucial for enriching group consensus learning and improving overall recommendation performance.
- The geometric centroid method generally proves more effective than the geometric barycenter for deriving common member preferences, particularly for group recommendation tasks.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.