Classification of small-ball modes and maximum a posteriori estimators in metric spaces

TL;DR

This paper develops a systematic framework for defining modes in metric spaces based on small-ball probabilities, classifies ten consistent definitions, and explores their properties and relationships.

Contribution

It introduces a comprehensive axiomatic framework for modes in metric spaces, classifies all consistent definitions, and analyzes their structural relationships.

Findings

Exactly ten consistent mode definitions identified

Modes form a complete, distributive lattice structure

Classification simplifies for well-behaved measures

Abstract

A mode, or `most likely point', for a probability measure $\mu$ can be defined in various ways via the asymptotic behaviour of the $\mu$-mass of balls as their radius tends to zero. Such points are of intrinsic interest in the local theory of measures on metric spaces and also arise naturally in the study of Bayesian inverse problems and diffusion processes. Building upon special cases already proposed in the literature, this paper develops a systematic framework for defining modes through small-ball probabilities. We propose `common-sense' axioms that such definitions should obey, including appropriate treatment of discrete and absolutely continuous measures, as well as symmetry and invariance properties. We show that there are exactly ten such definitions consistent with these axioms, and that they are partially but not totally ordered in strength, forming a complete, distributive lattice. We also show how this classification simplifies for well-behaved $\mu$.