Research map / Learning theory and algorithms / Matrix and tensor methods
Fast Randomized Singular Value Thresholding for Low-Rank Optimization
Tae-Hyun Oh, Yasuyuki Matsushita, Yu‐Wing Tai, In So Kweon
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 exact economy SVD with FRSVT-RP in inexact ALM RPCA | iALM RPCA · RPCA 2000x2000 | total time (s): 2751.910 -> 5.123 (537.17x speed-up) | Improved |
| replace exact economy SVD with FRSVT-RP in inexact ALM RPCA | iALM RPCA · RPCA 4000x4000 | total time (s): 21668.280 -> 30.274 (715.74x speed-up) | Improved |
| replace exact SVD with FRSVT-RP | iALM RPCA · RPCA 10000x100 | total time (s): 3.217 -> 0.782 (4.11x speed-up) | Improved |
| replace exact SVD with FRSVT-RP | iALM RPCA · RPCA 6000x6000 | NRMSE: 1.42e-07 -> 1.43e-07 | No clear effect |
| add range propagation (RP) to FRSVT | iALM-FRSVT · RPCA 2000x2000 | total time (s): 7.928 -> 5.123 | Improved |
| add adaptive rank prediction (AP) to fixed over-sampling p=2 (ablation) | iALM-FRSVT and iALM-RSVD · RPCA convergence | convergence to ground truth: does not converge -> converges | Improved |
| increase number of power iterations beyond one (ablation) | iALM-FRSVT-WS · RPCA convergence | NRMSE per iteration | No clear effect |
| use randomly projected nuclear norm (Mu et al.) instead of exact ALM SVT | RPCA · RPCA 2000x2000 | NRMSE: 2.11e-07 -> 1.10e+00 | Hurt |
| replace SVD with FRSVT in LRR | robust LRR · Hopkins155 motion segmentation | time per motion (s): 1.016 -> 0.452 | Improved |
| replace SVD with FRSVT in LRR | robust LRR · Extended Yale B | computation time (s): 75.216 -> 44.590 | Improved |
| replace SVD with FRSVT in LRR | robust LRR · Hopkins155 motion segmentation | segmentation error (%): 1.59 -> 1.59 | No clear effect |
| replace SVD with FRSVT in LRR | robust LRR · Extended Yale B | segmentation accuracy (%): 79.06 -> 79.06 | No clear effect |
Closest papers by meaning
- A generalization of the randomized singular value decomposition (ICLR 2022 · similarity 0.87)
- Large-Scale Low-Rank Matrix Learning with Nonconvex Regularizers (TPAMI 2018 · similarity 0.86)
- Fast Stochastic Algorithms for SVD and PCA: Convergence Properties and Convexity (ICML 2016 · similarity 0.85)
- Iteratively Capped Reweighting Norm Minimization With Global Convergence Guarantee for Low-Rank Matrix Learning (TPAMI 2024 · similarity 0.84)
- Solving L1-regularized SVMs and Related Linear Programs: Revisiting the Effectiveness of Column and Constraint Generation (JMLR 2022 · similarity 0.84)
- OTLRM: Orthogonal Learning-based Low-Rank Metric for Multi-Dimensional Inverse Problems (AAAI 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.
