Work

Martingale posteriors for generative classifiers

Year: 2025

Type: article

Abstract: Generative models for classification are a well-established method in statistics and machine learning. Martingales posteriors provide a computationally feasible method for performing prior-free Bayesian analysis. This paper aims to address the problem of uncertainty quantification through martingale posteriors for generative models for classification. To this aim, a conditionally identically distributed sequence of observations is considered. An empirical analysis is given.

Source: Statistics & Probability Letters

DOI: 10.1016/j.spl.2025.110627

Authors: Pier Giovanni Bissiri, Matteo Borrotti

Institution: University of Milano-Bicocca

Language: English

Topic: Statistical Methods and Bayesian Inference

Subfield: Statistics and Probability

Field: Mathematics

Domain: Physical Sciences

Open Access status: hybrid

Estimated APC paid: $3,180

Funder: Università degli Studi di Milano-Bicocca