Research map / Learning theory and algorithms / Kernels and regression theory
Towards a Unified Quadrature Framework for Large-Scale Kernel Machines
Fanghui Liu, Xiaolin Huang, Yudong Chen, Johan A. K. Suykens
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 fifth-degree D-FS instead of SGQ | deterministic kernel approximation · magic04, letter, ijcnn1, covtype | number of nodes and approximation error: 221 -> 201 nodes on magic04 | Improved |
| use fifth-degree D-FS instead of SGQ | deterministic kernel approximation · covtype | number of nodes: 5941 -> 5833 | Improved |
| use third-degree D-FS instead of RFF and QMC | Gaussian kernel approximation · four datasets | approximation error | Improved |
| use third-degree D-FS instead of SSR | Gaussian kernel approximation · four datasets | approximation error | Hurt |
| use third-degree D-FS instead of SGQ | Gaussian kernel approximation · four datasets | approximation error | No clear effect |
| use fifth-degree D-FS instead of quadrature methods SSR and SGQ | Gaussian kernel approximation · four datasets | time cost | Improved |
| use quadrature approximation methods instead of RFF on prediction | KRR classification · four datasets | test accuracy | No clear effect |
| add control variates (S-FS+RFF) to RFF (ablation) | random Fourier features · magic04, Gaussian kernel | approximation error | Improved |
| add control variates (S-FS+RFF) to RFF (ablation) | random Fourier features · magic04, Gaussian kernel | time cost | Hurt |
| increase feature dimension D for S-FS versus RFF | S-FS · Gaussian kernel | variance reduction | Hurt |
| use first-order arc-cosine kernel instead of Gaussian kernel | kernel approximation · four datasets | approximation error | Hurt |
| add QMC sampling (S-FS+QMC) instead of SSR sampling | S-FS · ijcnn1 and covtype | time cost | Improved |
Closest papers by meaning
- Quadrature-based features for kernel approximation (NeurIPS 2018 · similarity 0.91)
- Convergence guarantees for kernel-based quadrature rules in misspecified settings (NeurIPS 2016 · similarity 0.89)
- On the Equivalence between Kernel Quadrature Rules and Random Feature Expansions (JMLR 2017 · similarity 0.89)
- Kernel quadrature with DPPs (NeurIPS 2019 · similarity 0.88)
- Fast Summation of Radial Kernels via QMC Slicing (ICLR 2025 · similarity 0.86)
- Quasi-Monte Carlo Features for Kernel Approximation (ICML 2024 · similarity 0.85)
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.
