Research map / Learning theory and algorithms / Matrix and tensor methods
Truncated Cauchy Non-Negative Matrix Factorization
Naiyang Guan, Tongliang Liu, Yangmuzi Zhang, Dacheng Tao, Larry Steven Davis
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 truncation to Cauchy loss (Truncated CauchyNMF vs CauchyNMF) (ablation) | CauchyNMF · ORL occlusion (block b=16) | clustering accuracy (%): 18.36 -> 47.30 | Improved |
| add truncation to Cauchy loss (ablation) | CauchyNMF · ORL occlusion (block b=10) | clustering accuracy (%): 19.27 -> 57.80 | Improved |
| use Truncated CauchyNMF instead of CIM-NMF under heavy occlusion | CIM-NMF · ORL occlusion (block b=14) | clustering accuracy (%): 26.88 -> 55.38 | Improved |
| use Truncated CauchyNMF instead of CIM-NMF at mild occlusion | CIM-NMF · ORL occlusion (block b=10) | clustering accuracy (%): 58.48 -> 57.80 | No clear effect |
| use Truncated CauchyNMF instead of L1-NMF under Laplace noise | L1-NMF · ORL Laplace noise (delta=200) | relative reconstruction error (%): 29.73 -> 18.10 | Improved |
| use Truncated CauchyNMF instead of L2-NMF under Laplace noise | L2-NMF · ORL Laplace noise (delta=280) | relative reconstruction error (%): 82.06 -> 27.23 | Improved |
| use Truncated CauchyNMF instead of L1-NMF under Salt & Pepper noise (40%) | L1-NMF · ORL Salt & Pepper noise (40% pixels) | relative reconstruction error (%): 16.75 -> 12.35 | Improved |
| use Truncated CauchyNMF under 60% Salt & Pepper corruption | L1-NMF · ORL Salt & Pepper noise (60% pixels) | relative reconstruction error (%): 29.62 -> 35.26 | Hurt |
| use Truncated CauchyNMF at 5% Salt & Pepper corruption | L1-NMF · ORL Salt & Pepper noise (5% pixels) | relative reconstruction error (%): 11.33 -> 12.37 | No clear effect |
| use Truncated CauchyNMF instead of SEC for AR face recognition with sunglasses | SEC · AR face recognition (sunglasses) | accuracy (%): 51.67 -> 77 | Improved |
| use Truncated CauchyNMF+SRC instead of L1-NMF+SRC | L1-NMF+SRC · AR face recognition (total) | accuracy (%): 87.15 -> 90 | Improved |
| use Truncated CauchyNMF on synthetic line data with 4/9 outliers | L1-NMF · synthetic 2D subspace | subspace recovery | Improved |
Closest papers by meaning
- Robust Matrix Factorization by Majorization Minimization (TPAMI 2017 · similarity 0.89)
- Non-negative Matrix Factorization under Heavy Noise (ICML 2016 · similarity 0.88)
- Sparse NMF with Archetypal Regularization: Computational and Robustness Properties (JMLR 2024 · similarity 0.88)
- Distributionally Robust and Multi-Objective Nonnegative Matrix Factorization (TPAMI 2021 · similarity 0.88)
- Subspace Clustering With Priors via Sparse Quadratically Constrained Quadratic Programming (CVPR 2016 · similarity 0.88)
- RGNMR: A Gauss-Newton method for robust matrix completion with theoretical guarantees (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.
