Research map / Learning theory and algorithms / Matrix and tensor methods
Triple Component Matrix Factorization: Untangling Global, Local, and Noisy Components
Naichen Shi, Salar Fattahi, Raed Al Kontar
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 least-squares joint/individual factorization (JIVE, HMF) with TCMF that models sparse noise | TCMF · Synthetic data (alpha=0.01) | log10 global error: 5.49 -> -3.38 | Improved |
| TCMF vs best baseline at higher sparsity alpha=0.1 | TCMF · Synthetic data (alpha=0.1) | log10 global error: 6.48 -> -1.93 | Improved |
| increase noise sparsity parameter alpha from 0.01 to 0.1 (ablation) | TCMF · Synthetic data | log10 global error: -3.38 -> -1.93 | Hurt |
| model sparse noise explicitly (TCMF vs HMF) | TCMF · Noisy surveillance video (background) | MSE: 198 -> 0.0003 | Improved |
| model sparse noise explicitly (TCMF vs RaJIVE) | TCMF · Noisy surveillance video (foreground) | MSE: 166 -> 98 | Improved |
| add joint/individual decomposition on top of robust PCA (TCMF vs Robust PCA) | TCMF · Noisy surveillance video (foreground) | MSE: 5105 -> 98 | Improved |
| TCMF vs Robust PCA on background recovery | TCMF · Noisy surveillance video (background) | MSE: 0.0016 -> 0.0003 | Improved |
| TCMF vs RaJIVE on running time | TCMF · Noisy surveillance video | wall-clock time (s): 1.2×10^4 -> 3.5×10^2 | Improved |
| TCMF vs Robust PCA on running time | TCMF · Noisy surveillance video | wall-clock time (s): 3.3×10^-1 -> 3.5×10^2 | Hurt |
Closest papers by meaning
- Structured Low-Rank Matrix Factorization: Global Optimality, Algorithms, and Applications (TPAMI 2019 · similarity 0.88)
- Truncated Cauchy Non-Negative Matrix Factorization (TPAMI 2017 · similarity 0.86)
- The Incremental Multiresolution Matrix Factorization Algorithm (CVPR 2017 · similarity 0.86)
- Provable Alternating Gradient Descent for Non-negative Matrix Factorization with Strong Correlations (ICML 2017 · similarity 0.85)
- Compact Matrix Factorization With Dependent Subspaces (CVPR 2017 · similarity 0.85)
- Deep Rank-One Tensor Functional Factorization for Multi-Dimensional Data Recovery (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.
