Reproduce “Represent Items by Items: An Enhanced Representation of the Target Item for 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
Yinjiang Cai, Zeyu Cui, Shu Wu, Zhen Lei, Xibo Ma
Researchers from the Chinese Academy of Sciences developed Co-occurrence based Enhanced Representation (CER) to improve target item representations in recommender systems by leveraging collaborative signals from co-occurring items. This approach, which is particularly effective for low-frequency items, improved HR@10 by 17.87% and NDCG@10 by 26.74% on the Amazon-Books dataset over existing baselines.
Key insights
- The quality of target item representations, particularly for low-frequency items, constitutes a critical and often overlooked bottleneck in Item-based Collaborative Filtering.
- Leveraging collaborative signals from co-occurring items can significantly enrich the semantic information of target items, mitigating issues caused by sparse interaction data.
- Enhanced item representations can be pre-computed offline, ensuring efficient online inference and practical deployability for real-time recommendation systems.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.