Gurtej Kanwar (Massachusetts Institute of Technology)
Critical slowing down and topological freezing are key obstacles to progress in lattice QCD calculations of hadronic properties causing the cost of ensemble generation to severely diverge in the continuum limit. Recently, a class of machine learning techniques known as flow-based models has been successfully applied to produce exact sampling schemes that can circumvent critical slowing down and/or topological freezing in purely bosonic proof-of-principle applications. I will briefly summarize these flow-based MCMC methods and discuss progress towards including the contributions of fermionic degrees of freedom in this method, required for example to include dynamical quark contributions to flow-based sampling for lattice QCD.
Michael Albergo (NYU) Gurtej Kanwar (Massachusetts Institute of Technology) Racanière Sébastien (DeepMind) Danilo Jimenez Rezende (DeepMind) Julian M. Urban (University of Heidelberg) Dr Denis Boyda (MIT / Argonne National Lab) Prof. Kyle Cranmer (NYU) Dr Daniel C. Hackett (MIT) Phiala Shanahan (MIT)