Properties and limitations of geometric tempering for gradient flow dynamics

TL;DR

This paper analyzes the properties and limitations of geometric tempering in gradient flow dynamics for sampling, providing convergence bounds and exploring discretizations and adaptive schedules.

Contribution

It offers new theoretical bounds on convergence for tempered gradient flows and introduces novel adaptive tempering schedules.

Findings

Exponential convergence in continuous time for tempered gradient flows.

Discrete-time discretizations have specific convergence properties.

Geometric mixture of initial and target distributions does not speed up convergence.

Abstract

We consider the problem of sampling from a probability distribution $\pi$. It is well known that this can be written as an optimisation problem over the space of probability distributions in which we aim to minimise the Kullback--Leibler divergence from $\pi$. We consider the effect of replacing $\pi$ with a sequence of moving targets $(\pi_t)_{t\ge0}$ defined via geometric tempering on the Wasserstein and Fisher--Rao gradient flows. We show that convergence occurs exponentially in continuous time, providing novel bounds in both cases. We also consider popular time discretisations and explore their convergence properties. We show that in the Fisher--Rao case, replacing the target distribution with a geometric mixture of initial and target distribution never leads to a convergence speed up both in continuous time and in discrete time. Finally, we explore the gradient flow structure of tempered dynamics and derive novel adaptive tempering schedules.