Research map / Learning theory and algorithms / Matrix and tensor methods
Sparse PCA: a Geometric Approach
Dimitris Bertsimas, Driss Lahlou Kitane
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 GeoSPCA (all PCs at once) instead of greedy | GeoSPCA · Arcene (a=3, k=14) | explained variance: 8.2E+8 -> 1.02E+9 (+24.2%) | Improved |
| build all PCs at once (GeoSPCA) instead of deflation (ablation) | GeoSPCA · Mturk (a=8, k=16) | explained variance: 5.62E+3 -> 1.59E+5 | Improved |
| build all PCs at once (GeoSPCA) instead of deflation (ablation) | GeoSPCA · Arcene (a=3, k=14) | explained variance: 7.43E+7 -> 1.02E+9 | Improved |
| build all PCs at once (GeoSPCA) instead of deflation (ablation) | GeoSPCA · CGD (a=11, k=23) | explained variance: 5.11E+10 -> 1.89E+12 | Improved |
| GeoSPCA vs greedy as support size k grows | GeoSPCA · Arcene (a=3) | relative variance gain over Greedy: +24.2% at k=14 -> +0.4% at k=270 | Mixed |
| GeoSPCA vs greedy as support size k grows | GeoSPCA · Colon (a=5) | relative variance gain over Greedy: +4.9% at k=11 -> -0.3% at k=33 | Mixed |
| GeoSPCA vs greedy as support size k grows | GeoSPCA · Mturk (a=8) | relative variance gain over Greedy: +1.1% at k=16 -> +0.4% at k=32 | Improved |
| GeoSPCA vs greedy on CGD | GeoSPCA · CGD (a=11) | relative variance gain over Greedy: +2.7%, +5.1%, +3.9%, +2.9%, +0.6% | Improved |
| use GeoSPCA instead of PathSPCA/Greedy for computation time | GeoSPCA · CGD | computation time: PathSPCA 2 mins to 11 hours; Greedy 4 mins to 51 hours; GeoSPCA did not exceed 2 mins | Improved |
| use GeoSPCA instead of Greedy for computation time | GeoSPCA · Arcene | computation time: Greedy 2min to 8 hours; PathSPCA 4 to 8 mins; GeoSPCA <1s to 45s | Improved |
| use GeoSSPCA instead of Structured Sparse PCA for face recognition | GeoSSPCA · Face recognition (occluded test images) | precision: more than 10% improvement | Improved |
Closest papers by meaning
- Efficient Sparse PCA via Block-Diagonalization (ICLR 2025 · similarity 0.91)
- Solving Large-Scale Sparse PCA to Certifiable (Near) Optimality (JMLR 2022 · similarity 0.91)
- Upper bounds for Model-Free Row-Sparse Principal Component Analysis (ICML 2020 · similarity 0.90)
- Fast and Provable Algorithms for Sparse PCA with Improved Sample Complexity (ICML 2025 · similarity 0.87)
- Oja's Algorithm for Streaming Sparse PCA (NeurIPS 2024 · similarity 0.87)
- Rethinking PCA Through Duality (NeurIPS 2025 · similarity 0.87)
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.
