Research map / Learning theory and algorithms / Matrix and tensor methods
Dimension-free Concentration Bounds on Hankel Matrices for Spectral Learning
François Denis, Mattias Gybels, Amaury Habrard
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 dimension-free Theorem 7 instead of Eq. (3) | concentration bound · PAutomaC problem 3 | bound value: 0.100 -> 0.067 | Improved |
| compute sigma^2 over U x V (opt.) rather than Sigma* x Sigma* (ablation) | Theorem 7 · PAutomaC problem 15 | bound value: 0.085 -> 0.028 | Improved |
| use Theorem 7 instead of Eq. (3) | Hankel bound · PAutomaC problem 15 | bound value: 0.127 -> 0.085 | Improved |
| use Theorem 9 instead of Eq. (3), prefix case eta=1/2 | prefix Hankel bound · PAutomaC problem 3 | bound value: 0.140 -> 0.089 | Improved |
| compute sigma^2 over U x V for Theorem 9, eta=1/2 (ablation) | Theorem 9 · PAutomaC problem 3 | bound value: 0.089 -> 0.064 | Improved |
| use Theorem 9 (opt.) instead of Eq. (3), eta=1 | prefix Hankel bound · PAutomaC problem 3 | bound value: 0.281 -> 0.128 | Improved |
| use Theorem 11 instead of Eq. (3), factor case | factor Hankel bound · PAutomaC problem 3 | bound value: 0.148 -> 0.099 | Improved |
| compute sigma^2 over U x V for Theorem 11 (ablation) | Theorem 11 · PAutomaC problem 3 | bound value: 0.099 -> 0.060 | Improved |
| use factor variant instead of prefix variant of Hankel matrix | spectral learning · PAutomaC (11 problems) | distance between singular-vector spaces | Improved |
| use prefix variant instead of classical Hankel matrix | spectral learning · PAutomaC (11 problems) | distance between singular-vector spaces | Improved |
| increase eta from 1/2 to 1 in prefix/factor variants | spectral learning · PAutomaC (11 problems) | distance between singular-vector spaces | Improved |
| increase |V| from 100 to 20,000 | spectral learning · PAutomaC (11 problems) | distance between singular-vector spaces | Hurt |
Closest papers by meaning
- Spectral Estimation with Free Decompression (NeurIPS 2025 · similarity 0.87)
- Intrinsic-dimension empirical Bernstein inequalities for bounded self-adjoint operators (NeurIPS 2026 · similarity 0.85)
- Free Decompression with Algebraic Spectral Curves (NeurIPS 2026 · similarity 0.85)
- Optimal Rates for Vector-Valued Spectral Regularization Learning Algorithms (NeurIPS 2024 · similarity 0.85)
- A Unified Framework for Learning with Nonlinear Model Classes from Arbitrary Linear Samples (ICML 2024 · similarity 0.84)
- Spectral Perturbation Bounds for Low-Rank Approximation with Applications to Privacy (NeurIPS 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.
