Testing algorithms for critical slowing down

INSPIRE: 1631653
arXiv: 1710.07036
DOI: 10.1051/epjconf/201817502008

Authors: Cossu, Guido, Boyle, Peter, Christ, Norman, Jung, Chulwoo, Jüttner, Andreas, Sanfilippo, Francesco

Submitted: 19 October 2017

Subjects:

• hep-lat
• Lattice
• Computing

Journal reference: EPJ Web Conf. 175 02008 (2018)

Citations: 16

Abstract

We present the preliminary tests on two modifications of the Hybrid Monte Carlo (HMC) algorithm. Both algorithms are designed to travel much farther in the Hamiltonian phase space for each trajectory and reduce the autocorrelations among physical observables thus tackling the critical slowing down towards the continuum limit. We present a comparison of costs of the new algorithms with the standard HMC evolution for pure gauge fields, studying the autocorrelation times for various quantities including the topological charge.