Research map / Optimization / Nonconvex and smooth optimization
Approximate Secular Equations for the Cubic Regularization Subproblem
Yihang Gao, Man-Chung Yue, Michael Ng
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 ARC-ASEM(1) instead of ARC-Krylov(10) | ARC subproblem solver · TOINTGSS | gradient norm: 2.20E-08 -> 8.01E-10 | Improved |
| Use ARC-ASEM(1) instead of ARC-Krylov(30) | ARC subproblem solver · BRYBAND | gradient norm: 1.14E-07 -> 1.02E-07 | Improved |
| Use ARC-ASEM(10) instead of ARC-Krylov(30) | ARC subproblem solver · DIXMAANG | gradient norm: 9.06E-09 -> 4.85E-09 | Improved |
| Use ARC-ASEM(1) instead of ARC-Krylov(10) | ARC subproblem solver · TQUARTIC | gradient norm: 8.48E-09 -> 9.62E-09 | Hurt |
| Increase the eigenvalue count from 1 to 10 (ablation) | ARC-ASEM · TOINTGSS | gradient norm: 8.01E-10 -> 8.17E-10 | Hurt |
| Increase the eigenvalue count from 1 to 10 (ablation) | ARC-ASEM · BRYBAND | gradient norm: 1.02E-07 -> 1.01E-07 | Improved |
| Increase the eigenvalue count from 1 to 10 (ablation) | ARC-ASEM · DIXMAANG | gradient norm: 5.53E-09 -> 4.85E-09 | Improved |
| Increase the eigenvalue count from 1 to 10 (ablation) | ARC-ASEM · TQUARTIC | gradient norm: 9.62E-09 -> 9.62E-09 | No clear effect |
| Use the second-order truncated secular equation rather than the first-order version (ablation) | ASEM · synthetic cubic regularization subproblem | convergence speed | Improved |
| Use an unsuitable or random value of μ rather than a suitable estimate (ablation) | ASEM · synthetic cubic regularization subproblem | convergence | Hurt |
| Use an m-dimensional Krylov subspace to estimate the eigenvalues and eigenvectors (ablation) | ASEM · synthetic cubic regularization subproblem | suboptimality | No clear effect |
| Use first-order ASEM with clustered unobserved eigenvalues rather than widely spread ones (ablation) | ASEM · synthetic cubic regularization subproblem | convergence speed | Improved |
Closest papers by meaning
- Accelerating Adaptive Cubic Regularization of Newton's Method via Random Sampling (JMLR 2022 · similarity 0.90)
- Convergence of Cubic Regularization for Nonconvex Optimization under KL Property (NeurIPS 2018 · similarity 0.88)
- Cauchy-Schwarz Regularizers (ICLR 2025 · similarity 0.85)
- Spectral Preconditioning for Gradient Methods on Graded Non-convex Functions (ICML 2024 · similarity 0.83)
- A Regularized Newton Method for Nonconvex Optimization with Global and Local Complexity Guarantees (NeurIPS 2025 · similarity 0.83)
- AdaCubic: An Adaptive Cubic Regularization Optimizer for Deep Learning (NeurIPS 2026 · 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.
