Minimax Confidence Intervals for the Sliced Wasserstein Distance

Abstract

Motivated by the growing popularity of variants of the Wasserstein distance in statistics and machine learning, we study statistical inference for the Sliced Wasserstein distance--an easily computable variant of the Wasserstein distance. Specifically, we construct confidence intervals for the Sliced Wasserstein distance which have finite-sample validity under no assumptions or under mild moment assumptions. These intervals are adaptive in length to the regularity of the underlying distributions. We also bound the minimax risk of estimating the Sliced Wasserstein distance, and as a consequence establish that the lengths of our proposed confidence intervals are minimax optimal over appropriate distribution classes. To motivate the choice of these classes, we also study minimax rates of estimating a distribution under the Sliced Wasserstein distance. These theoretical findings are complemented with a simulation study demonstrating the deficiencies of the classical bootstrap, and the advantages of our proposed methods. We also show strong correspondences between our theoretical predictions and the adaptivity of our confidence interval lengths in simulations. We conclude by demonstrating the use of our confidence intervals in the setting of simulator-based likelihood-free inference. In this setting, contrasting popular approximate Bayesian computation methods, we develop uncertainty quantification methods with rigorous frequentist coverage guarantees.

arXiv commentPublished at https://doi.org/10.1214/22-EJS2001 in the Electronic Journal of Statistics

References in corpus (25)

• Statistical Aspects of Wasserstein Distances
• Mining gold from implicit models to improve likelihood-free inference
• Approximate Bayesian computation with the Wasserstein distance
• Concentration inequalities and asymptotic results for ratio type empirical processes
• Convergence of latent mixing measures in finite and infinite mixture models
• Convergence and Concentration of Empirical Measures under Wasserstein Distance in Unbounded Functional Spaces
• Subspace Robust Wasserstein Distances
• Sequential Monte Carlo with Adaptive Weights for Approximate Bayesian Computation
• Multilevel Clustering via Wasserstein Means
• K2-ABC: Approximate Bayesian Computation with Kernel Embeddings
• Minimax Distribution Estimation in Wasserstein Distance
• Likelihood-free inference with an improved cross-entropy estimator
• Generalized Sliced Wasserstein Distances
• Convergence Rates for Gaussian Mixtures of Experts
• A Unifying Approach to Distributional Limits for Empirical Optimal Transport
• Central Limit Theorems for General Transportation Costs
• On the Minimax Optimality of Estimating the Wasserstein Metric
• Distributional Sliced-Wasserstein and Applications to Generative Modeling
• Limit Laws for Empirical Optimal Solutions in Stochastic Linear Programs
• Likelihood-free parameter estimation for dynamic queueing networks: case study of passenger flow in an international airport terminal
• Validation of Approximate Likelihood and Emulator Models for Computationally Intensive Simulations
• Confidence Sets and Hypothesis Testing in a Likelihood-Free Inference Setting
• Tree-Sliced Variants of Wasserstein Distances
• A Central Limit Theorem for $L_p$ transportation cost with applications to Fairness Assessment in Machine Learning
• Hypothesis Test and Confidence Analysis with Wasserstein Distance on General Dimension

Cited by in corpus (7)

• Faster Wasserstein Distance Estimation with the Sinkhorn Divergence
• Sharp Convergence Rates for Empirical Optimal Transport with Smooth Costs
• Statistical and Topological Properties of Sliced Probability Divergences
• Two-sample Test using Projected Wasserstein Distance
• Martingale Methods for Sequential Estimation of Convex Functionals and Divergences
• On Projection Robust Optimal Transport: Sample Complexity and Model Misspecification
• Uniform Confidence Band for Optimal Transport Map on One-Dimensional Data