Reproduce “Rethinking Missing Data: Aleatoric Uncertainty-Aware 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
Chenxu Wang, Fuli Feng, Yang Zhang, Qifan Wang, Xunhan Hu, Xiangnan He
Researchers at the University of Science and Technology of China and Meta AI introduce an Aleatoric Uncertainty-aware Recommendation (AUR) framework that models the inherent data randomness in missing implicit feedback. This approach robustly enhances recommendations for long-tail items by over 50% and maintains overall accuracy across various backbone models.
Key insights
- Missing implicit feedback data should be understood as inherently uncertain rather than definitively negative, with this aleatoric uncertainty providing a valuable signal.
- Explicitly modeling and incorporating fine-grained, user-item specific aleatoric uncertainty into recommendation scores effectively mitigates the mislabeling issue.
- Utilizing uncertainty, especially with a higher tail-controlling coefficient, disproportionately benefits long-tail items by compensating for their higher likelihood of mislabeling without sacrificing overall performance.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.