Research map / Label-efficient and robust learning / Recognition and long-tail
Smooth neighborhood recommender systems
Ben Dai, Junhui Wang, Xiaotong Shen, Annie Qu
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 |
|---|---|---|---|
| add user-item network smoothing (sSVD) over group-specific SVD | gSVD · Last.fm (entire pairs) | RMSE: 1.026 -> 0.860 | Improved |
| add user-item network smoothing (sSVD) over group-specific SVD | gSVD · Last.fm (cold-start pairs) | RMSE: 1.202 -> 0.706 | Improved |
| add neighborhood smoothing with social network and tags over rSVD | rSVD · Last.fm (entire pairs) | RMSE: 1.237 -> 0.860 | Improved |
| add network and noise-robust smoothing over rSVD | rSVD · Simulation K=6, missing 0.95, cold-start 0.2 | RMSE: 1.154 -> 0.460 | Improved |
| add user-item network smoothing over gSVD under high cold-start | gSVD · Simulation K=6, missing 0.95, cold-start 0.2 | RMSE: 0.866 -> 0.460 | Improved |
| add user-item network smoothing at no cold-start and low missing | gSVD · Simulation K=3, missing 0.8, cold-start 0.0 | RMSE: 0.302 -> 0.364 | Hurt |
| add the full network (vs removing all edges) (ablation) | sSVD · Simulation K=6, missing 0.95, noise 0.1, cold-start 0.0 | RMSE: 0.713 -> 0.396 | Improved |
| remove half of the network edges (ablation) | sSVD · Simulation K=6, missing 0.95, noise 0.1, cold-start 0.0 | RMSE: 0.396 -> 0.569 | Hurt |
| increase neighborhood noise from 0.1 to 0.6 (ablation) | sSVD · Simulation K=6, missing 0.95, full network | RMSE: 0.396 -> 1.547 | Hurt |
| add the network under high noise (vs removing it) (ablation) | sSVD · Simulation K=6, missing 0.95, noise 0.6, cold-start 0.0 | RMSE: 1.397 -> 1.547 | Hurt |
| increase missing rate from 0.8 to 0.95 at cold-start 0.2 (ablation) | sSVD · Simulation K=3 | RMSE: 0.375 -> 0.440 | Mixed |
| increase missing rate from 0.8 to 0.95 at cold-start 0.2 (ablation) | gSVD · Simulation K=3 | RMSE: 0.578 -> 0.849 | Hurt |
Closest papers by meaning
- Social Collaborative Filtering by Trust (TPAMI 2016 · similarity 0.89)
- Recommendations with Sparse Comparison Data: Provably Fast Convergence for Nonconvex Matrix Factorization (ICML 2025 · similarity 0.88)
- Be Aware of the Neighborhood Effect: Modeling Selection Bias under Interference (ICLR 2024 · similarity 0.86)
- Collaborative Filtering With Social Exposure: A Modular Approach to Social Recommendation (AAAI 2018 · similarity 0.86)
- 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 0.85)
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.
