The 38th International Symposium on Lattice Field Theory

Neural Network Field Transformation and Its Application in HMC

29 Jul 2021, 13:45

15m

Oral presentation Algorithms (including Machine Learning, Quantum Computing, Tensor Networks) Algorithms (including Machine Learning, Quantum Computing, Tensor Networks)

Speaker

Xiao-Yong Jin (Argonne National Laboratory)

Description

We propose a continuous differentiable and invertible gauge field transformation parametrized with neural networks. We apply this technique to 2D U(1) pure gauge system, combine the transformation and HMC, and train the neural network field transformation for improved tunneling of topological sectors during the gauge generation. We present the properties of the trained transformation applied to the gauge fields with different gauge couplings and lattice volumes, and discuss the cost and benefit of this algorithm in tackling the critical slowing down of generating independent gauge configurations toward the continuum limit.

Presentation materials

613_R3Al_Xiao-Yong_Jin

talk.pdf