Research map / Learning theory and algorithms / Matrix and tensor methods
Hessian Eigenspectra of More Realistic Nonlinear Models
Zhenyu Liao, Michael W. Mahoney
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 logistic loss instead of exponential loss for the logistic model | G-GLM Hessian · Logistic model | eigenvalue support | Improved |
| Use exponential loss instead of logistic loss for the logistic model | G-GLM Hessian · Logistic model | eigenvalue support | Hurt |
| Use logistic loss with identity covariance | G-GLM Hessian · Logistic model | similarity to Marchenko–Pastur law | Improved |
| Replace identity covariance with heterogeneous feature covariance | G-GLM Hessian · Logistic model | number of spectral bulks | Mixed |
| Use the phase retrieval response model with square loss | G-GLM Hessian · Phase retrieval | eigenvalue support | Hurt |
| Add nonzero data-mean signal | Logistic model Hessian · Logistic model, C = I_p | alignment of top eigenvector with μ: ρ > √c | Improved |
| Increase the model parameter norm ‖w‖ | Logistic model Hessian · Logistic model, μ = 0 | eigengap of model-induced spike | Mixed |
| Increase the model parameter norm ‖w‖ | Logistic model Hessian · Logistic model, μ = 0 | alignment of isolated eigenvector with w | Improved |
| Apply the proposed preprocessing function to phase retrieval measurements | Phase retrieval spectral initialization · Phase retrieval, w proportional to w* | right-side isolated spike | Improved |
Closest papers by meaning
- Investigating the Overlooked Hessian Structure: From CNNs to LLMs (ICML 2025 · similarity 0.84)
- Free Decompression with Algebraic Spectral Curves (NeurIPS 2026 · similarity 0.83)
- Neglected Hessian component explains mysteries in sharpness regularization (NeurIPS 2024 · similarity 0.83)
- Learning single index models via harmonic decomposition (NeurIPS 2025 · similarity 0.83)
- Nonlinear Laplacians: Tunable principal component analysis under directional prior information (NeurIPS 2025 · similarity 0.82)
- Spectral Estimation with Deformed Decompression (NeurIPS 2026 · similarity 0.82)
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.
