Research map / Learning theory and algorithms / Prediction and decision losses
Collaborative Filtering With Social Exposure: A Modular Approach to Social Recommendation
Menghan Wang, Xiaolin Zheng, Yang Yang, Kun Zhang
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Use social information to model item exposure, then condition collaborative filtering on exposure | SERec · Lastfm, Delicious, Douban, Epinions | Recall@50 | Improved |
| Use social boosting rather than social regularization for the exposure component | SERec · Lastfm, Douban, Epinions | Recall@50: 0.3990 -> 0.4381; 0.2302 -> 0.2425; 0.1347 -> 0.1556 | Improved |
| Use social boosting rather than social regularization for the exposure component | SERec · Delicious | Recall@50: 0.3927 -> 0.4442 | Hurt |
| Add social exposure modeling using social boosting instead of ExpoMF's non-social exposure | ExpoMF · Delicious | Recall@50: 0.4448 -> 0.4442 | Hurt |
| Add social exposure modeling using social boosting instead of ExpoMF's non-social exposure | ExpoMF · Lastfm, Douban, Epinions | NDCG@100: 0.2200 -> 0.3102; 0.1840 -> 0.2015; 0.0634 -> 0.0777 | Improved |
| Reduce the social links used to 60% or 20% (ablation) | SERec_regular · Lastfm and Delicious | Recall@10: Lastfm: 0.2110 -> 0.2071 / 0.1782; Delicious: 0.1385 -> 0.0802 / 0.0701 | Hurt |
| Reduce the social links used to 60% or 20% (ablation) | SERec_boost · Lastfm and Delicious | Recall@10: Lastfm: 0.2142 -> 0.1701 / 0.1553; Delicious: 0.1934 -> 0.1693 / 0.1426 | Hurt |
| Reduce the social links used to 60% or 20% (ablation) | SERec_regular · Lastfm and Delicious | Recall@50: Lastfm: 0.3971 -> 0.3910 / 0.3233; Delicious: 0.3927 -> 0.2398 / 0.2020 | Hurt |
| Reduce the social links used to 60% or 20% (ablation) | SERec_boost · Lastfm and Delicious | Recall@50: Lastfm: 0.4386 -> 0.3933 / 0.3801; Delicious: 0.4442 -> 0.3866 / 0.3739 | Hurt |
| Use social boosting rather than social regularization as social links are pruned (ablation) | SERec · Lastfm and Delicious | Recall@50: Lastfm decay ratio: 18.58% vs 13.34%; Delicious: 48.56% vs 15.83% | Mixed |
Closest papers by meaning
- Social Recommendation with an Essential Preference Space (AAAI 2018 · similarity 0.89)
- Fairly Recommending with Social Attributes: A Flexible and Controllable Optimization Approach (NeurIPS 2023 · similarity 0.88)
- Fading to Grow: Growing Preference Ratios via Preference Fading Discrete Diffusion for Recommendation (NeurIPS 2025 · similarity 0.86)
- Would You Like Your Data to Be Trained? A User Controllable Recommendation Framework (AAAI 2024 · similarity 0.86)
- Optimizing Social Network Interventions via Hypergradient-Based Recommender System Design (ICML 2025 · similarity 0.86)
- From Pairwise to Ranking: Climbing the Ladder to Ideal Collaborative Filtering with Pseudo-Ranking (AAAI 2025 · similarity 0.85)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
