Research map / Learning theory and algorithms / Matrix and tensor methods
Unlabeled Principal Component Analysis and Matrix Completion
Yunzhen Yao, Liangzu Peng, Manolis C. Tsakiris
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 |
|---|---|---|---|
| use UPCA pipeline with LSRF in Stage-II instead of observed corrupted data | UPCA on Breast Cancer Wisconsin · Breast Cancer Wisconsin (alpha=0.7) | relative estimation error: 32.24% to 6.35% | Improved |
| use LSRF instead of l1-RR or PL in Stage-II | UPCA with DPCP · Breast Cancer Wisconsin (alpha=0.7) | relative estimation error: 15.90% and 19.57% vs 6.35% | Improved |
| use LSRF instead of l1-RR in Stage-II | UPCA synthetic · Synthetic, 90% outliers | runtime: 1.5min to 0.3sec | Improved |
| use estimated subspace from DPCP instead of ground-truth subspace (ablation) | UPCA Stage-II methods · Synthetic, 90% outliers | relative estimation error | No clear effect |
| decrease permutation ratio alpha to 0.1 for DPCP and OP (ablation) | Stage-I robust PCA · Synthetic Stage-I | principal angle: alpha=0.1, r=25: OP 7, DPCP (10^-6) degrees | Improved |
| decrease permutation ratio alpha to 0.1 for CoP and Self-Expr (ablation) | Stage-I robust PCA · Synthetic Stage-I | principal angle: 67, 80 degrees at alpha=0.1, r=25 | Hurt |
| use DPCP+IST instead of MCO with more missing entries | UMC Stage-I · Synthetic UMC, 50% outliers, 50% missing | principal angle: 20.37 to 2.61 | Improved |
| use DPCP+IST instead of MCO with few missing entries | UMC Stage-I · Synthetic UMC, 50% outliers, 10% missing | principal angle: 2.44 vs 5.51 | Hurt |
| apply UMC algorithm to restore high-school scores | UMC · High-school scores, 25% missing | relative error: 51% to 17% | Improved |
| apply UMC algorithm to restore breast tumor features | UMC · Breast tumor, 25% missing | relative error: 51% to 37% | Improved |
| use CCV-Min instead of AIEM | UPCA Stage-II · Synthetic, r=3, 75% outliers | error and runtime: 1% error in 42msec vs 0.42% in 15sec | Mixed |
Closest papers by meaning
- Unlabeled Principal Component Analysis (NeurIPS 2021 · similarity 0.98)
- Multi-Subspace Matrix Recovery from Permuted Data (AAAI 2025 · similarity 0.88)
- RES-PCA: A Scalable Approach to Recovering Low-Rank Matrices (CVPR 2019 · similarity 0.88)
- Coherence Pursuit: Fast, Simple, and Robust Subspace Recovery (ICML 2017 · similarity 0.87)
- Learned Robust PCA: A Scalable Deep Unfolding Approach for High-Dimensional Outlier Detection (NeurIPS 2021 · similarity 0.87)
- Fun with Flags: Robust Principal Directions via Flag Manifolds (CVPR 2024 · similarity 0.85)
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.
