Characteristic and Universal Tensor Product Kernels

Abstract

Maximum mean discrepancy (MMD), also called energy distance or N-distance in statistics and Hilbert-Schmidt independence criterion (HSIC), specifically distance covariance in statistics, are among the most popular and successful approaches to quantify the difference and independence of random variables, respectively. Thanks to their kernel-based foundations, MMD and HSIC are applicable on a wide variety of domains. Despite their tremendous success, quite little is known about when HSIC characterizes independence and when MMD with tensor product kernel can discriminate probability distributions. In this paper, we answer these questions by studying various notions of characteristic property of the tensor product kernel.

final version appeared in JMLR

References in corpus (1)

• Kernel dimension reduction in regression

Cited by in corpus (16)

• Gaussian Processes and Kernel Methods: A Review on Connections and Equivalences
• Information Constraints on Auto-Encoding Variational Bayes
• Measuring Association on Topological Spaces Using Kernels and Geometric Graphs
• A Measure-Theoretic Approach to Kernel Conditional Mean Embeddings
• DP-MERF: Differentially Private Mean Embeddings with Random Features for Practical Privacy-Preserving Data Generation
• Self-Supervised Learning with Kernel Dependence Maximization
• Kernel Partial Correlation Coefficient -- a Measure of Conditional Dependence
• On Distance and Kernel Measures of Conditional Independence
• Hard Shape-Constrained Kernel Machines
• Testing Goodness of Fit of Conditional Density Models with Kernels
• Consistency of permutation tests for HSIC and dHSIC
• Calibration tests beyond classification
• Bandits for Learning to Explain from Explanations
• Dynamic Flows on Curved Space Generated by Labeled Data
• A Kernel Two-sample Test for Dynamical Systems
• Estimating Rényi's $\alpha$-Cross-Entropies in a Matrix-Based Way