PIRSA:19070080

Self-learning Monte Carlo method and structured self-attention network

Liu, J. (2019). Self-learning Monte Carlo method and structured self-attention network. Perimeter Institute. https://pirsa.org/19070080

Liu, Junwei. Self-learning Monte Carlo method and structured self-attention network. Perimeter Institute, Jul. 16, 2019, https://pirsa.org/19070080 

          @misc{ pirsa_PIRSA:19070080,
            doi = {10.48660/19070080},
            url = {https://pirsa.org/19070080},
            author = {Liu, Junwei},
            keywords = {Condensed Matter},
            language = {en},
            title = {Self-learning Monte Carlo method and structured self-attention network},
            publisher = {Perimeter Institute},
            year = {2019},
            month = {jul},
            note = {PIRSA:19070080 see, \url{https://pirsa.org}}
          }

Junwei Liu

Hong Kong University of Science and Technology

Talk number
PIRSA:19070080
Collection
Talk Type
Subject
Abstract

Self-learning Monte Carlo (SLMC) method is a general-purpose numerical method to simulate many-body systems. SLMC can efficiently cure the critical slowing down in both bosonic and fermionic systems. Moreover, for fermionic systems, SLMC can generally reduce the computational complexity and speed up simulations even away from the critical points. For example, SLMC is more than 1000 times faster than the conventional method for the double exchange model in 8*8*8 cubic lattice. In addition, SLMC also provides a general framework to naturally integrate the advanced machine learning techniques into Monte Carlo. In this talk, I will give an introduction about the background, basic idea and the design principle of SLMC. Later, I will explicitly show how to use SLMC and its great accelerations in classic systems, free fermions coupled with classical spins systems, and interacting fermion systems. At the end, I will talk our recent developed a new type of neural network, which is motivated by the mean field theory and could be used to represent many physical Hamiltonians. We believe the new structural network would be highly efficient to identify different phases and accelerate the numerical simulations for even complex models.