Research map / Learning theory and algorithms / Matrix and tensor methods
An $\ell_{\infty}$ Eigenvector Perturbation Bound and Its Application
Jianqing Fan, Weichen Wang, Yiqiao Zhong
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 |
|---|---|---|---|
| increase matrix size d (d from 200 to 2000) (ablation) | l_inf eigenvector perturbation error · Simulation r=3 | max-entry error err: slope around -0.5 in log-log | Improved |
| increase eigengap gamma with gamma*sqrt(d) fixed (ablation) | l_inf eigenvector perturbation error · Simulation r=3 | err x gamma sqrt(d) | No clear effect |
| change perturbation generating mechanism (sparse uniform vs autoregressive structure) (ablation) | l_inf eigenvector perturbation error · Simulation r=3, ||E||_inf about 30 | err | No clear effect |
| use robust Huber-based estimator instead of sample covariance in POET | POET covariance estimation · Elliptical t, df=3 | error ratio vs sample covariance | Improved |
| use robust Huber-based estimator instead of sample covariance in POET | POET covariance estimation · Gaussian data | error ratio vs sample covariance | No clear effect |
| use Kendall's tau instead of sample covariance in POET | POET covariance estimation · Gaussian data | median error ratio | Hurt |
| use spatial Kendall's tau instead of marginal Kendall's tau eigenvectors | POET covariance estimation · Elliptical t | robustness | Improved |
| use Huber-based robust estimator instead of Kendall's tau methods | POET covariance estimation · Element-wise iid t (non-elliptical) | error ratio | Improved |
Closest papers by meaning
- Spectral Perturbation Bounds for Low-Rank Approximation with Applications to Privacy (NeurIPS 2025 · similarity 0.88)
- Coherence-free Entrywise Estimation of Eigenvectors in Low-rank Signal-plus-noise Matrix Models (NeurIPS 2024 · similarity 0.86)
- Perturbation Bounds for Low-Rank Inverse Approximations under Noise (NeurIPS 2025 · similarity 0.86)
- Singular Subspace Perturbation Bounds via Rectangular Random Matrix Diffusions (ICLR 2025 · similarity 0.85)
- Entrywise error bounds for low-rank approximations of kernel matrices (NeurIPS 2024 · similarity 0.84)
- Regularized least squares learning with heavy-tailed noise is minimax optimal (NeurIPS 2025 · similarity 0.83)
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.
