Temperature-steerable flows
Computational Physics
2020-12-02 v1
Abstract
Boltzmann generators approach the sampling problem in many-body physics by combining a normalizing flow and a statistical reweighting method to generate samples of a physical system's equilibrium density. The equilibrium distribution is usually defined by an energy function and a thermodynamic state, such as a given temperature. Here we propose temperature-steerable flows (TSF) which are able to generate a family of probability densities parametrized by a choosable temperature parameter. TSFs can be embedded in a generalized ensemble sampling framework such as parallel tempering in order to sample a physical system across thermodynamic states, such as multiple temperatures.
Cite
@article{arxiv.2012.00429,
title = {Temperature-steerable flows},
author = {Manuel Dibak and Leon Klein and Frank Noé},
journal= {arXiv preprint arXiv:2012.00429},
year = {2020}
}