Research map / Learning theory and algorithms / Kernels and regression theory
Separation Results between Fixed-Kernel and Feature-Learning Probability Metrics
Carles Domingo-Enrich, Youssef Mroueh
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 feature-learning F1 discriminators instead of fixed-kernel F2 discriminators | Probability metrics on constructed hypersphere distribution pairs · Hypersphere pairs μ_d, ν_d with high-frequency Legendre-harmonic differences | IPM: d_{B_F1}(μ_d,ν_d) / d_{B_F2}(μ_d,ν_d) = √N_{k,d} | Improved |
| Use feature-learning F1 Stein discriminators instead of fixed-kernel F2 Stein discriminators | Stein discrepancy on constructed hypersphere distribution pairs · Hypersphere pairs μ_d, ν_d with high-frequency Legendre-harmonic differences | Stein discrepancy: SD_{B_F1^d}(μ_d,ν_d) = 1 | Improved |
| Use feature-learning F1 rather than fixed-kernel F2 | Probability metrics comparing distributions that differ in one Gaussian variance · Gaussian distributions, varying dimensions | discrimination by IPM: F1 IPM is constant; F2 IPM decreases | Improved |
| Use max-sliced Wasserstein instead of sliced Wasserstein | Gaussian distribution comparison · Gaussian distributions, varying dimensions | discrimination distance: max-sliced Wasserstein distance is constant; sliced Wasserstein distance decreases | Improved |
| Use feature-learning F1 rather than fixed-kernel F2 Stein discrepancy | Stein discrepancy estimates · Hypersphere pairs μ_d, ν_d with k=5 | ratio of F1 to F2 Stein discrepancies: empirical ratio is significantly above the theoretical lower bound | Improved |
Closest papers by meaning
- Kernel Distribution Embeddings: Universal Kernels, Characteristic Kernels and Kernel Metrics on Distributions (JMLR 2018 · similarity 0.86)
- Kernel Quantile Embeddings and Associated Probability Metrics (ICML 2025 · similarity 0.84)
- The optimality of kernel classifiers in Sobolev space (ICLR 2024 · similarity 0.84)
- Stabilizing Sample Similarity in Representation via Mitigating Random Consistency (ICML 2025 · similarity 0.84)
- Learning to Embed Distributions via Maximum Kernel Entropy (NeurIPS 2024 · similarity 0.84)
- From Kernels to Features: A Multi-Scale Adaptive Theory of Feature Learning (ICML 2025 · similarity 0.83)
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.
