Kernel Mean Embedding of Distributions: A Review and Beyond

Abstract

A Hilbert space embedding of a distribution---in short, a kernel mean embedding---has recently emerged as a powerful tool for machine learning and inference. The basic idea behind this framework is to map distributions into a reproducing kernel Hilbert space (RKHS) in which the whole arsenal of kernel methods can be extended to probability measures. It can be viewed as a generalization of the original "feature map" common to support vector machines (SVMs) and other kernel methods. While initially closely associated with the latter, it has meanwhile found application in fields ranging from kernel machines and probabilistic modeling to statistical inference, causal discovery, and deep learning. The goal of this survey is to give a comprehensive review of existing work and recent advances in this research area, and to discuss the most challenging issues and open problems that could lead to new research directions. The survey begins with a brief introduction to the RKHS and positive definite kernels which forms the backbone of this survey, followed by a thorough discussion of the Hilbert space embedding of marginal distributions, theoretical guarantees, and a review of its applications. The embedding of distributions enables us to apply RKHS methods to probability measures which prompts a wide range of applications such as kernel two-sample testing, independent testing, and learning on distributional data. Next, we discuss the Hilbert space embedding for conditional distributions, give theoretical insights, and review some applications. The conditional mean embedding enables us to perform sum, product, and Bayes' rules---which are ubiquitous in graphical model, probabilistic inference, and reinforcement learning---in a non-parametric way. We then discuss relationships between this framework and other related areas. Lastly, we give some suggestions on future research directions.

147 pages; this is the final version

References in corpus (23)

• Learning Transferable Features with Deep Adaptation Networks
• Brownian distance covariance
• Domain Generalization via Invariant Feature Representation
• Generative Moment Matching Networks
• Kernel-based Conditional Independence Test and Application in Causal Discovery
• UAV-Assisted Heterogeneous Networks for Capacity Enhancement
• Convolutional Neural Network Architectures for Signals Supported on Graphs
• Kernel dimension reduction in regression
• Stability Properties of Graph Neural Networks
• Training generative neural networks via Maximum Mean Discrepancy optimization
• Learning from Distributions via Support Measure Machines
• Kernel Belief Propagation
• Nonparametric Divergence Estimation with Applications to Machine Learning on Distributions
• Testing for Homogeneity with Kernel Fisher Discriminant Analysis
• Relational Forward Models for Multi-Agent Learning
• FastMMD: Ensemble of Circular Discrepancy for Efficient Two-Sample Test
• Sequential Kernel Herding: Frank-Wolfe Optimization for Particle Filtering
• Herding Dynamic Weights for Partially Observed Random Field Models
• Convergence guarantees for kernel-based quadrature rules in misspecified settings
• Hilbert Space Embeddings of POMDPs
• Detecting low-complexity unobserved causes
• Kernel Mean Estimation via Spectral Filtering
• A low variance consistent test of relative dependency

Cited by in corpus (57)

• MMD GAN: Towards Deeper Understanding of Moment Matching Network
• Deep transfer operator learning for partial differential equations under conditional shift
• Eigendecompositions of Transfer Operators in Reproducing Kernel Hilbert Spaces
• Kernel-based approximation of the Koopman generator and Schrödinger operator
• Comparison of Maximum Likelihood and GAN-based training of Real NVPs
• Robust Domain Misinformation Detection via Multi-modal Feature Alignment
• Discovering and Deciphering Relationships Across Disparate Data Modalities
• A kernel-based approach to molecular conformation analysis
• Kernel methods for detecting coherent structures in dynamical data
• Anomaly Detection Based on Isolation Mechanisms: A Survey
• Distribution-based Sketching of Single-Cell Samples
• Finite sample properties of parametric MMD estimation: robustness to misspecification and dependence
• Learnware: Small Models Do Big
• Kernel method for persistence diagrams via kernel embedding and weight factor
• Nonequilibrium Statistical Mechanics and Optimal Prediction of Partially-Observed Complex Systems
• Supervised Learning with Quantum Measurements
• Quantum Mean Embedding of Probability Distributions
• Automatic and effective discovery of quantum kernels
• Methods for Quantifying Dataset Similarity: a Review, Taxonomy and Comparison
• Transparent Single-Cell Set Classification with Kernel Mean Embeddings
• Local Group Invariant Representations via Orbit Embeddings
• Change Detection of Markov Kernels with Unknown Pre and Post Change Kernel
• Singular Value Decomposition of Operators on Reproducing Kernel Hilbert Spaces
• Feature-Based Interpolation and Geodesics in the Latent Spaces of Generative Models
• Steganographer Identification
• Geometric Methods for Sampling, Optimisation, Inference and Adaptive Agents
• Distribution regression model with a Reproducing Kernel Hilbert Space approach
• Detecting Change Intervals with Isolation Distributional Kernel
• When Invariant Representation Learning Meets Label Shift: Insufficiency and Theoretical Insights
• Multimodal Multi-User Surface Recognition with the Kernel Two-Sample Test
• Estimation of copulas via Maximum Mean Discrepancy
• Data-Driven Observability Analysis for Nonlinear Stochastic Systems
• Confidence-based Estimators for Predictive Performance in Model Monitoring
• Diffeological statistical models, the Fisher metric and probabilistic mappings
• Testing and Learning on Distributions with Symmetric Noise Invariance
• Linear quadratic control of nonlinear systems with Koopman operator learning and the Nyström method
• Koopman operators with intrinsic observables in rigged reproducing kernel Hilbert spaces
• Learning Interpretable Collective Variables for Spreading Processes on Networks
• A Distributional Treatment of Real2Sim2Real for Object-Centric Agent Adaptation in Vision-Driven Deformable Linear Object Manipulation
• Approximating Symmetrized Estimators of Scatter via Balanced Incomplete U-Statistics
• Towards a Kernel based Uncertainty Decomposition Framework for Data and Models
• Permutation Invariant Policy Optimization for Mean-Field Multi-Agent Reinforcement Learning: A Principled Approach
• Ergodic Trajectory Optimization on Generalized Domains Using Maximum Mean Discrepancy
• Approximate Bayesian Computation Based on Maxima Weighted Isolation Kernel Mapping
• Causal survival embeddings: non-parametric counterfactual inference under censoring
• Model-Free Change Point Detection for Mixing Processes
• Fully symmetric kernel quadrature
• Reversible Gromov-Monge Sampler for Simulation-Based Inference
• FastPart: Over-Parameterized Stochastic Gradient Descent for Sparse optimisation on Measures
• Revisiting RIP guarantees for sketching operators on mixture models
• Signal reconstruction using determinantal sampling
• Instrument Space Selection for Kernel Maximum Moment Restriction
• Generalization Analysis of Transformers in Distribution Regression
• Composable Learning with Sparse Kernel Representations
• Minimum Discrepancy Methods in Uncertainty Quantification
• Hyperparameter Learning for Conditional Kernel Mean Embeddings with Rademacher Complexity Bounds
• Nonlinear Distribution Regression for Remote Sensing Applications