Research map / Learning theory and algorithms / Kernels and regression theory
Linear-Time Learning on Distributions with Approximate Kernel Embeddings
Danica Sutherland, Junier Oliva, Barnabás Póczos, Jeff Schneider
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 |
|---|---|---|---|
| approximate JS kernel with random features (A + RKS) instead of direct entropy-based estimation | JS RBF kernel · Gram matrix estimation, 50 GMMs | R2 with true kernel: 0.981 -> 0.966 | No clear effect |
| add RKS embedding on top of projection-coefficient features (ablation) | JS kernel features A · Gram matrix estimation, 50 GMMs | R2 with true kernel: 0.974 -> 0.966 | Hurt |
| use HDD (JS, Hellinger, TV) random-feature kernels instead of L2 and MMD kernels | ridge regression · Mixture component count | RMSE | Improved |
| use HDD kernels instead of histogram JS embedding | ridge regression · Mixture component count, |chi|=200 | RMSE | Improved |
| use TV embedding instead of histogram JS embedding | SVM on distributions · CIFAR-10 cat vs dog | misclassification rate | Improved |
| use JS or Hellinger embedding instead of histogram JS | SVM on distributions · CIFAR-10 cat vs dog | misclassification rate | No clear effect |
| use HDD embeddings instead of D3, L2 and histogram embeddings | VGG-16 conv features · Scene-15 | accuracy | Improved |
| use HDD random features instead of the full-Gram KL kernel | SVM on distributions · CIFAR-10 cat vs dog | computation time | Improved |
Closest papers by meaning
- Kernel Distribution Embeddings: Universal Kernels, Characteristic Kernels and Kernel Metrics on Distributions (JMLR 2018 · similarity 0.88)
- Kernel Quantile Embeddings and Associated Probability Metrics (ICML 2025 · similarity 0.87)
- Learning to Embed Distributions via Maximum Kernel Entropy (NeurIPS 2024 · similarity 0.87)
- On Statistical Learning Theory for Distributional Inputs (ICML 2024 · similarity 0.86)
- On the Consistency of Kernel Methods with Dependent Observations (ICML 2024 · similarity 0.86)
- Random features for Grassmannian kernel approximation with bounded rank-one projections (NeurIPS 2026 · 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.
