Huembeli, Patrick ORCID: 0000-0001-7047-6897, Dauphin, Alexandre ORCID: 0000-0003-4996-2561, Wittek, Peter and Gogolin, Christian ORCID: 0000-0003-0290-4698 (2019). Automated discovery of characteristic features of phase transitions in many-body localization. Phys. Rev. B, 99 (10). COLLEGE PK: AMER PHYSICAL SOC. ISSN 2469-9969

Full text not available from this repository.

Abstract

We identify a new order parameter for the disorder-driven many-body localization transition by leveraging machine learning. Contrary to previous studies, our method is almost entirely unsupervised. A game theoretic process between neural networks defines an adversarial setup with conflicting objectives to identify what characteristic features to base efficient predictions on. This reduces the numerical effort for mapping out the phase diagram by a factor of 100x and allows us to pin down the transition, as the point at which the physics changes qualitatively, in an objective and cleaner way than is possible with the existing diverse array of quantities. Our approach of automated discovery is applicable specifically to poorly understood phase transitions and is a starting point for a research program leveraging the potential of machine learning-assisted research in physics.

Item Type: Journal Article
Creators:
CreatorsEmailORCIDORCID Put Code
Huembeli, PatrickUNSPECIFIEDorcid.org/0000-0001-7047-6897UNSPECIFIED
Dauphin, AlexandreUNSPECIFIEDorcid.org/0000-0003-4996-2561UNSPECIFIED
Wittek, PeterUNSPECIFIEDUNSPECIFIEDUNSPECIFIED
Gogolin, ChristianUNSPECIFIEDorcid.org/0000-0003-0290-4698UNSPECIFIED
URN: urn:nbn:de:hbz:38-153454
DOI: 10.1103/PhysRevB.99.104106
Journal or Publication Title: Phys. Rev. B
Volume: 99
Number: 10
Date: 2019
Publisher: AMER PHYSICAL SOC
Place of Publication: COLLEGE PK
ISSN: 2469-9969
Language: English
Faculty: Unspecified
Divisions: Unspecified
Subjects: no entry
Uncontrolled Keywords:
KeywordsLanguage
Materials Science, Multidisciplinary; Physics, Applied; Physics, Condensed MatterMultiple languages
Refereed: Yes
URI: http://kups.ub.uni-koeln.de/id/eprint/15345

Downloads

Downloads per month over past year

Altmetric

Export

Actions (login required)

View Item View Item