%0 Journal Article %1 wirnsberger2020targeted %A Wirnsberger, Peter %A Ballard, Andrew J. %A Papamakarios, George %A Abercrombie, Stuart %A Racanière, Sébastien %A Pritzel, Alexander %A Rezende, Danilo Jimenez %A Blundell, Charles %D 2020 %K energy-models generative-models %T Targeted free energy estimation via learned mappings %U http://arxiv.org/abs/2002.04913 %X Free energy perturbation (FEP) was proposed by Zwanzig more than six decades ago as a method to estimate free energy differences, and has since inspired a huge body of related methods that use it as an integral building block. Being an importance sampling based estimator, however, FEP suffers from a severe limitation: the requirement of sufficient overlap between distributions. One strategy to mitigate this problem, called Targeted Free Energy Perturbation, uses a high-dimensional mapping in configuration space to increase overlap of the underlying distributions. Despite its potential, this method has attracted only limited attention due to the formidable challenge of formulating a tractable mapping. Here, we cast Targeted FEP as a machine learning (ML) problem in which the mapping is parameterized as a neural network that is optimized so as to increase overlap. We test our method on a fully-periodic solvation system, with a model that respects the inherent permutational and periodic symmetries of the problem. We demonstrate that our method leads to a substantial variance reduction in free energy estimates when compared against baselines.

@article{wirnsberger2020targeted, abstract = {Free energy perturbation (FEP) was proposed by Zwanzig more than six decades ago as a method to estimate free energy differences, and has since inspired a huge body of related methods that use it as an integral building block. Being an importance sampling based estimator, however, FEP suffers from a severe limitation: the requirement of sufficient overlap between distributions. One strategy to mitigate this problem, called Targeted Free Energy Perturbation, uses a high-dimensional mapping in configuration space to increase overlap of the underlying distributions. Despite its potential, this method has attracted only limited attention due to the formidable challenge of formulating a tractable mapping. Here, we cast Targeted FEP as a machine learning (ML) problem in which the mapping is parameterized as a neural network that is optimized so as to increase overlap. We test our method on a fully-periodic solvation system, with a model that respects the inherent permutational and periodic symmetries of the problem. We demonstrate that our method leads to a substantial variance reduction in free energy estimates when compared against baselines.}, added-at = {2020-02-13T11:57:02.000+0100}, author = {Wirnsberger, Peter and Ballard, Andrew J. and Papamakarios, George and Abercrombie, Stuart and Racanière, Sébastien and Pritzel, Alexander and Rezende, Danilo Jimenez and Blundell, Charles}, biburl = {https://www.bibsonomy.org/bibtex/244b4c2d0e679ffb3dfcbd89f24533363/kirk86}, description = {[2002.04913] Targeted free energy estimation via learned mappings}, interhash = {9036f98675653e6641819f2a2655ccb3}, intrahash = {44b4c2d0e679ffb3dfcbd89f24533363}, keywords = {energy-models generative-models}, note = {cite arxiv:2002.04913Comment: 10 pages, 5 figures}, timestamp = {2020-02-13T11:57:02.000+0100}, title = {Targeted free energy estimation via learned mappings}, url = {http://arxiv.org/abs/2002.04913}, year = 2020 }