Research map / Learning theory and algorithms / Kernels and regression theory
Learning Adaptive Random Features
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 |
|---|---|---|---|
| learn spectral samples and weights via EKAL (SAMPLE) instead of random MC sampling | random Fourier features · Wine r=200 | relative kernel approximation error: 0.13 -> 0.05 | Helped |
| learn spectral samples and weights via EKAL (SAMPLE) | random Fourier features · Covtype r=200 | relative kernel approximation error: 0.11 -> 0.05 | Helped |
| learn features via EKAL (SAMPLE) | random Fourier features · Adult r=200 | relative kernel approximation error: 0.14 -> 0.10 | Helped |
| learn weights only (WEIGHT) vs jointly learning samples and weights (SAMPLE) (ablation) | learned Fourier features · Parkinson r=200 | relative kernel approximation error: 0.03 -> 0.02 | Helped |
| use k-means cluster centers as landmarks instead of random landmarks (ablation) | EKAL Fourier features · Covtype r=200 | relative kernel approximation error: 0.05 -> 0.04 | Helped |
| use k-means cluster centers as landmarks instead of random landmarks (ablation) | EKAL Fourier features · MNIST r=200 | relative kernel approximation error: 0.06 -> 0.10 | Hurt |
| increase number of landmarks n from r/25 up to r (ablation) | SAMPLE and CLUSTER · Wine, Parkinson | relative error | Helped |
| add supervised hybrid loss with EKAL regularization (SUPERVISE vs SAMPLE) (ablation) | learned Fourier features · CPU | RMSE: 7.495 -> 3.687 | Helped |
| add supervised hybrid loss with EKAL regularization (SUPERVISE vs SAMPLE) (ablation) | learned Fourier features · MNIST | error (%): 0.757 -> 0.378 | Helped |
| add supervised hybrid loss with EKAL regularization (SUPERVISE vs SAMPLE) (ablation) | learned Fourier features · Parkinson | RMSE: 6.957 -> 6.432 | Helped |
| add hybrid supervised loss (SUPERVISE) vs ALIGN supervised baseline | learned Fourier features · Adult | error (%): 15.94 -> 15.64 | Helped |
| unsupervised EKAL features (SAMPLE) vs MC features | random Fourier features · regression/classification benchmarks | RMSE / error | No clear effect |
Closest papers by meaning
- Towards a Unified Analysis of Random Fourier Features (ICML 2019 · similarity 0.91)
- TRF: Learning Kernels with Tuned Random Features (AAAI 2022 · similarity 0.91)
- Sharp Analysis of Random Fourier Features in Classification (AAAI 2022 · similarity 0.91)
- Random Fourier Features via Fast Surrogate Leverage Weighted Sampling (AAAI 2020 · similarity 0.90)
- Random Fourier Features for Kernel Ridge Regression: Approximation Bounds and Statistical Guarantees (ICML 2017 · similarity 0.90)
- Efficiently Learning Significant Fourier Feature Pairs for Statistical Independence Testing (NeurIPS 2024 · similarity 0.87)
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.