Köln 2025 – scientific programme
Parts | Days | Selection | Search | Updates | Downloads | Help
HK: Fachverband Physik der Hadronen und Kerne
HK 16: Heavy-Ion Collisions and QCD Phases III
HK 16.5: Talk
Tuesday, March 11, 2025, 15:00–15:15, HS 3 Chemie
Testing machine learning against finite size scaling in Lattice QCD — •Simran Singh1, Reinhold Kaiser2, Frithjof Karsch3, Jan Philipp Klinger2, Owe Philipsen2, and Christian Schmidt3 — 1HISKP Rheinischen Friedrich-Wilhelms Universität Bonn, Bonn, Germany — 2Institut für Theoretische Physik, Goethe-Universität Frankfurt, Frankfurt, Germany — 3Fakultät für Physik, Universität Bielefeld, Bielefeld, Germany
Masked Autoregressive Flows (MAFs) provide a machine learning method for estimating the joint probability density of observables from data samples. In [1], MAFs were used to estimate the joint probability density of the chiral condensate and gauge action conditioned on lattice parameters like gauge coupling, bare quark mass and spatial lattice extent for degenerate quarks using highly improved staggered fermions, identifying the critical mass separating crossover and first-order regions. This work extends the MAF analysis to previously published data using unimproved staggered fermions [2], aiming to compare MAF predictions of lattice observables with actual data with the ultimate goal to compare the ML approach to determine the Z2 critical mass with the finite size scaling analysis of the kurtosis, which was used in [3] by the Frankfurt group.
1. F. Karsch et.al., PoS LATTICE2022 (2023)
2. F. Cuteri et.al., JHEP 11 (2021)
3. O. Philipsen, PoS LATTICE2019 (2019)
Keywords: Chiral phase transition; Machine learning; QCD; lattice; finite size scaling