Research map / Learning theory and algorithms / Kernels and regression theory
Automated Spectral Kernel Learning
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 non-stationary instead of stationary kernels with assigned spectral density (ablation) | SK versus NSK · segment | accuracy (%): 89.93$\pm$2.12 -> 90.15$\pm$2.08 | Helped |
| Learn spectral density instead of assigning it, using a stationary kernel (ablation) | SK versus SKL · letter | accuracy (%): 76.50$\pm$1.21 -> 93.60$\pm$1.14 | Helped |
| Learn spectral density instead of assigning it, using a non-stationary kernel (ablation) | NSK versus NSKL · satimage | accuracy (%): 75.15$\pm$1.38 -> 83.74$\pm$1.34 | Helped |
| Add non-stationary kernels to a learned-spectrum method (ablation) | SKL versus NSKL · letter | accuracy (%): 93.60$\pm$1.14 -> 94.66$\pm$2.21 | Helped |
| Add trace-norm weight regularization and feature Frobenius-norm regularization (ablation) | ASKL versus NSKL · cadata | RMSE: 47.71$\pm$0.30 -> 46.34$\pm$0.23 | Helped |
| Add trace-norm weight regularization and feature Frobenius-norm regularization (ablation) | ASKL versus NSKL · abalone | RMSE: 7.85$\pm$0.42 -> 7.88$\pm$0.16 | Hurt |
| Add non-stationary kernels to a learned-spectrum method (ablation) | SKL versus NSKL · shuttle | accuracy (%): 98.87$\pm$1.42 -> 98.74$\pm$1.07 | Hurt |
| Use non-stationary kernels and learn the spectral density, with the proposed regularizers | ASKL · MNIST | accuracy (%): 96.67$\pm$1.61 -> 98.26$\pm$1.78 | Helped |
| Use non-stationary kernels and learn the spectral density, with the proposed regularizers | ASKL · USPS | accuracy (%): 95.27$\pm$1.65 -> 97.76$\pm$1.14 | Helped |
Closest papers by meaning
- Convolutional Spectral Kernel Learning with Generalization Guarantees (Abstract Reprint) (AAAI 2024 · similarity 0.90)
- Spectrum Dependent Learning Curves in Kernel Regression and Wide Neural Networks (ICML 2020 · similarity 0.90)
- Generalization error of spectral algorithms (ICLR 2024 · similarity 0.89)
- Optimal Rates for Vector-Valued Spectral Regularization Learning Algorithms (NeurIPS 2024 · similarity 0.88)
- Generalization in Kernel Regression Under Realistic Assumptions (ICML 2024 · similarity 0.88)
- Spectrally Transformed Kernel Regression (ICLR 2024 · similarity 0.86)
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.