Research map / Learning theory and algorithms / Matrix and tensor methods
Solving Large-Scale Sparse PCA to Certifiable (Near) Optimality
Dimitris Bertsimas, Ryan Cory-Wright, Jean Pauphilet
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 Gershgorin circle theorem bounds to the master problem (ablation) | cutting-plane algorithm · Communities (p=101), k=5 | runtime / solved: (2.87%) gap unsolved -> 0.20 s | Improved |
| add Gershgorin bounds to the master problem (ablation) | cutting-plane algorithm · Pitprops (p=13), k=5 | runtime (s): 0.30 -> 0.06 | Improved |
| add Gershgorin bounds to the master problem (ablation) | cutting-plane algorithm · Micromass (p=1300), k=5 | runtime (s): (1.29%) gap unsolved -> 163.60 | Improved |
| add Gershgorin bounds to the master problem (ablation) | cutting-plane algorithm · Micromass (p=1300), k=20 | bound gap: (35.9%) -> (35.9%) | No clear effect |
| add strengthening valid inequalities to the relaxation (ablation) | one-matrix relaxation with rounding · Pitprops (p=13), k=5 | relaxation gap (%): 23.8% -> 0.71% | Improved |
| add strengthening valid inequalities to the relaxation (ablation) | one-matrix relaxation with rounding · Miniboone (p=50), k=5 | relaxation gap (%): 781.3% -> 0.00% | Improved |
| add strengthening valid inequalities to the relaxation (ablation) | one-matrix relaxation with rounding · Miniboone (p=50), k=5 | objective gap (%): 235.6% -> 0.00% | Improved |
| add 20 linear PSD inequalities to the 2x2 minors outer approximation (ablation) | relax+round · Arrhythmia (p=274), k=20 | bound gap (%): 8.87% -> 4.48% | Improved |
| add 20 linear PSD inequalities to the 2x2 minors outer approximation (ablation) | relax+round · Micromass (p=1300), k=20 | runtime (s): 983.5 -> 16,350 | Hurt |
| use one-matrix relaxation with inequalities instead of p+1 matrix SDO relaxation | relax+round · Wine (p=13), k=5 | objective gap (%): 15.03% -> 0.00% | Improved |
| use relax+round instead of greedy method | sparse PCA · spiked covariance (p=150) | support recovery AUC: 0.7561 -> 0.7593 | Improved |
| use relax+round instead of sorting | sparse PCA · spiked covariance (p=150) | support recovery AUC: 0.7028 -> 0.7593 | Improved |
Closest papers by meaning
- Efficient Sparse PCA via Block-Diagonalization (ICLR 2025 · similarity 0.93)
- Upper bounds for Model-Free Row-Sparse Principal Component Analysis (ICML 2020 · similarity 0.93)
- Sparse PCA via Covariance Thresholding (JMLR 2016 · similarity 0.91)
- Fast and Provable Algorithms for Sparse PCA with Improved Sample Complexity (ICML 2025 · similarity 0.90)
- Oja's Algorithm for Streaming Sparse PCA (NeurIPS 2024 · similarity 0.88)
- Rethinking PCA Through Duality (NeurIPS 2025 · similarity 0.88)
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.
