Towards reduction of autocorrelation in HMC by machine learning

• DOI: 10.48550/arxiv.1712.03893
• Publication Date: 2017-01-01
• Authors: Akinori Tanaka; Akio Tomiya
• Keywords: FOS: Computer and information sciences; High Energy Physics - Lattice; Statistics - Machine Learning; 0103 physical sciences; High Energy Physics - Lattice (hep-lat); FOS: Physical sciences; Machine Learning (stat.ML); Disordered Systems and Neural Networks (cond-mat.dis-nn); Condensed Matter - Disordered Systems and Neural Networks; 01 natural sciences

In this paper we propose new algorithm to reduce autocorrelation in Markov chain Monte-Carlo algorithms for euclidean field theories on the lattice. Our proposing is Hybrid (HMC) with restricted Boltzmann machine. We examine validity of by employing phi-fourth theory three dimension. observe reduction both symmetric and broken phase as well. provides consistent central values expectation action density one-point Green's function ones from original HMC within statistical error. On other hand, two-point functions have slight difference between one calculated our phase. Furthermore, near criticality, distribution differs HMC. discuss origin discrepancies its improvement.