Local moment matching: A unified methodology for symmetric functional estimation and distribution estimation under Wasserstein distance

Abstract

We present \emph{Local Moment Matching (LMM)}, a unified methodology for symmetric functional estimation and distribution estimation under Wasserstein distance. We construct an efficiently computable estimator that achieves the minimax rates in estimating the distribution up to permutation, and show that the plug-in approach of our unlabeled distribution estimator is "universal" in estimating symmetric functionals of discrete distributions. Instead of doing best polynomial approximation explicitly as in existing literature of functional estimation, the plug-in approach conducts polynomial approximation implicitly and attains the optimal sample complexity for the entropy, power sum and support size functionals.

Cited by in corpus (12)

• 2-Wasserstein Approximation via Restricted Convex Potentials with Application to Improved Training for GANs
• Polynomial methods in statistical inference: theory and practice
• The Bethe and Sinkhorn Permanents of Low Rank Matrices and Implications for Profile Maximum Likelihood
• Profile Entropy: A Fundamental Measure for the Learnability and Compressibility of Discrete Distributions
• On the High Accuracy Limitation of Adaptive Property Estimation
• Optimal Estimation of Change in a Population of Parameters
• Regularized Weighted Chebyshev Approximations for Support Estimation
• Extrapolating the profile of a finite population
• On the Competitive Analysis and High Accuracy Optimality of Profile Maximum Likelihood
• The Broad Optimality of Profile Maximum Likelihood
• Optimal Estimation of Schatten Norms of a rectangular Matrix
• Instance Based Approximations to Profile Maximum Likelihood