Research map / Learning theory and algorithms / Matrix and tensor methods
Low-Rank Matrix Approximation with Stability
Dongsheng Li, Chao Chen, Qin Lv, Junchi Yan, Li Shang, Stephen Chu
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 stability terms over hard-predictable subsets (SMA) instead of plain RSVD | RSVD · MovieLens 10M | RMSE: 0.8256 ± 0.0006 -> 0.7682 ± 0.0003 | Improved |
| add stability terms (SMA) instead of plain RSVD | RSVD · Netflix | RMSE: 0.8534 ± 0.0001 -> 0.8036 ± 0.0004 | Improved |
| SMA instead of best ensemble WEMAREC | WEMAREC · MovieLens 10M | RMSE: 0.7775 ± 0.0007 -> 0.7682 ± 0.0003 | Improved |
| SMA instead of best ensemble WEMAREC | WEMAREC · Netflix | RMSE: 0.8143 ± 0.0001 -> 0.8036 ± 0.0004 | Improved |
| increase number of subsets K from 1 to 5 (ablation) | SMA · MovieLens 10M and Netflix | RMSE | Improved |
| SMA with K from 1 to 5 vs seven baselines (ablation) | SMA · MovieLens 10M and Netflix | RMSE | Improved |
| increase rank r from 50 to 250 (ablation) | SMA · MovieLens 10M and Netflix | RMSE | Improved |
| increase rank r above 50 (ablation) | RSVD · MovieLens 10M and Netflix | RMSE | Hurt |
| reduce training set ratio from 80% to 20% (ablation) | SMA and four single methods · MovieLens 10M | RMSE | Hurt |
| SMA vs single methods under sparse data (r=50) (ablation) | SMA · MovieLens 10M | RMSE | Improved |
| SMA vs RSVD training/test gap over epochs (r=20, K=3) (ablation) | SMA · MovieLens 10M | train-test RMSE gap | Improved |
| increase rank r from 5 to 20 | RSVD · MovieLens 1M | stability and generalization error | Hurt |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- LLORMA — LLORMA: Local Low-Rank Matrix Approximation (JMLR 2016) · better on 2 of 2 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Learning-Based Low-Rank Approximations (NeurIPS 2019 · similarity 0.89)
- LLORMA: Local Low-Rank Matrix Approximation (JMLR 2016 · similarity 0.88)
- Few-Shot Data-Driven Algorithms for Low Rank Approximation (NeurIPS 2021 · similarity 0.87)
- Additive Error Guarantees for Weighted Low Rank Approximation (ICML 2021 · similarity 0.86)
- COALA: Numerically Stable and Efficient Framework for Context-Aware Low-Rank Approximation (NeurIPS 2025 · similarity 0.84)
- From Low Rank Gradient Subspace Stabilization to Low-Rank Weights: Observations, Theories, and Applications (ICML 2025 · similarity 0.83)
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.
