On variational inference and maximum likelihood estimation with the {\lambda}-exponential family

TL;DR

This paper introduces a comprehensive framework for variational inference and maximum likelihood estimation within the { extbackslash lambda}-exponential family, extending classical methods to more general, heavy-tailed distributions with new optimality conditions and algorithms.

Contribution

It provides new theoretical optimality conditions and proximal algorithms for inference and estimation in the { extbackslash lambda}-exponential family, generalizing existing exponential family results.

Findings

New sufficient optimality conditions for variational inference.

Novel characterizations of maximum likelihood solutions.

Algorithms effective on heavy-tailed distributions.

Abstract

The {\lambda}-exponential family has recently been proposed to generalize the exponential family. While the exponential family is well-understood and widely used, this it not the case of the {\lambda}-exponential family. However, many applications require models that are more general than the exponential family. In this work, we propose a theoretical and algorithmic framework to solve variational inference and maximum likelihood estimation problems over the {\lambda}-exponential family. We give new sufficient optimality conditions for variational inference problems. Our conditions take the form of generalized moment-matching conditions and generalize existing similar results for the exponential family. We exhibit novel characterizations of the solutions of maximum likelihood estimation problems, that recover optimality conditions in the case of the exponential family. For the resolution of both problems, we propose novel proximal-like algorithms that exploit the geometry underlying the {\lambda}-exponential family. These new theoretical and methodological insights are tested on numerical examples, showcasing their usefulness and interest, especially on heavy-tailed target distributions.