Seminar "Representation of Quantum Many-body System by Neural Networks" by Mr. Nobuyuki Yoshioka

Speaker : Nobuyuki Yoshioka

Present : Ph.D. candidate at Univ. of Tokyo, Katsura group.  Working on interdisciplinary field between cond-mat physics and machine learning.
2017 : MSc. at Univ. of Tokyo
2015 : BSc. at Univ. of Tokyo

Abstract

Due to its high expressive power and the well-established learning algorithms, the neural networks are often applied to various tasks including image recognition, machine translation, game playing, and so on. Condensed matter theorists have recently become aware that the neural network is also suited to describe the many-body physical states in both closed and open quantum systems [1, 2].

In this talk, I would first introduce the expression of the Boltzmann machine, a specific type of neural networks, and give an overview of its application as the wave function ansatz for the closed quantum spin systems. Subsequently, I would discuss the extension of the ansatz to represent open quantum systems efficiently.

[1] G. Carleo and M. Troyer, Science 355, 602 (2017)

[2] N. Yoshioka and R. Hamazaki, arXiv:1902.07006 (2019)