Research map / Learning theory and algorithms / Sample complexity
Discrete-Valued Latent Preference Matrix Estimation with Graph Side Information
Changhun Jo, Kangwook Lee
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 |
|---|---|---|---|
| Replace symmetric two-level model with discrete multi-level latent preference model plus multi-cluster SBM and proposed algorithm | latent preference estimation with graph side information · Epinions (preprocessed) real rating/real graph | MAE: 0.540 vs 0.567 for Ahn's | Improved |
| Use the Ahn et al. symmetric-level algorithm when level pair is asymmetric (p1=0.3, p2=0.62) | Ahn et al. (2018) algorithm · synthetic ratings with synthetic SBM graph | estimation error (max norm) | Hurt |
| Ahn's algorithm error increases with observation rate p on real-graph setting, unlike other methods | Ahn et al. (2018) algorithm · Facebook graph with synthetic ratings (p1,p2,p3)=(0.05,0.5,0.95) | MAE | Hurt |
| Increase number of clusters/levels: ablation of l_max from 1 to 2 for the proposed algorithm (ablation) | proposed algorithm · Facebook graph with synthetic ratings | MAE | Improved |
| Refinement of clusters (Stage 2-(ii)) with latent preference vectors | proposed algorithm · noisy SBM synthetic | estimation error (max norm) | No clear effect |
| Larger alpha (graph quality) with beta fixed: p* decreases from 0.118 to 0.081 (ablation) | proposed algorithm theory · alpha 0.26 vs 0.27, beta 0.23, gamma 0.25 | threshold p*_gamma: 0.118 -> 0.081 | Improved |
| Increase gamma (fraction of differing coordinates) from 0.25 to 0.5 (ablation) | proposed algorithm · alpha 0.26, beta 0.23, d=3 | estimation error vs p | Improved |
Closest papers by meaning
- Binary Rating Estimation with Graph Side Information (NeurIPS 2018 · similarity 0.89)
- Discrete Social Recommendation (AAAI 2019 · similarity 0.86)
- Negative Sampling From the Ground Up: A Redesign for Recommendation (ICML 2026 · similarity 0.86)
- Recommendations with Sparse Comparison Data: Provably Fast Convergence for Nonconvex Matrix Factorization (ICML 2025 · similarity 0.85)
- Fading to Grow: Growing Preference Ratios via Preference Fading Discrete Diffusion for Recommendation (NeurIPS 2025 · similarity 0.85)
- Preference Diffusion for Recommendation (ICLR 2025 · similarity 0.84)
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.
