Research map / Learning theory and algorithms / Kernels and regression theory
Improved Random Features for Dot Product Kernels
Jonas Wacker, Motonobu Kanagawa, Maurizio Filippone
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 complex-valued instead of real-valued TensorSRHT (ablation) | TensorSRHT polynomial sketch · EEG, CIFAR10 ResNet34, MNIST, Gisette (nonnegative inputs) | variance of kernel estimate: smaller variance in 100%, 100%, 97.8%, 100% of pairs | Improved |
| use larger polynomial degree p with complex sketches | complex TensorSRHT · EEG, CIFAR10 ResNet34, MNIST, Gisette | variance ratio complex/real | Improved |
| use complex instead of real Rademacher sketch (ablation) | Rademacher sketch · synthetic positive-quadrant vectors, d=8 and d=32 | mean squared error ratio | Improved |
| use the tree construction with complex Rademacher sketch nodes instead of the proposed sketch | complex Rademacher sketch · synthetic positive-quadrant vectors, degrees p=2,3 | mean squared error | Hurt |
Closest papers by meaning
- Random Features for Kernel Approximation: A Survey on Algorithms, Theory, and Beyond (TPAMI 2021 · similarity 0.91)
- Learning Kernels with Random Features (NeurIPS 2016 · similarity 0.90)
- Dense-Exponential Random Features: Sharp Positive Estimators of the Gaussian Kernel (NeurIPS 2023 · similarity 0.88)
- Manifold Random Features (NeurIPS 2026 · similarity 0.86)
- Random features for Grassmannian kernel approximation with bounded rank-one projections (NeurIPS 2026 · similarity 0.86)
- Variance-Reducing Couplings for Random Features (ICLR 2025 · 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.
