SciPost Submission Page

Continuous-mixture Autoregressive Networks for efficient variational calculation of many-body systems

by Lingxiao Wang, Yin Jiang, Lianyi He, Kai Zhou

Submission summary

As Contributors: Lingxiao Wang · Kai Zhou
Preprint link: scipost_202104_00012v1
Date submitted: 2021-04-08 21:57
Submitted by: Zhou, Kai
Submitted to: SciPost Physics
Academic field: Physics
  • Probability
  • Artificial Intelligence
  • Condensed Matter Physics - Computational
Approaches: Theoretical, Computational, Phenomenological


Multi channel deep autoregressive networks were developed for variational calculation of many-body systems with continuous spin degrees of freedom. We embed the 2D XY model into the continuous-mixture networks and rediscover the topologial KT phase transition with vortices characterizing the quasi-long range order being detected. By learning the microscopic probability distributions, the networks compute the free energy directly and find that free vortices and anti-vortices emerge in the high-temperature regime.

Current status:
Editor-in-charge assigned

Submission & Refereeing History

You are currently on this page

Submission scipost_202104_00012v1 on 8 April 2021

Login to report or comment