mbl: Many-body localization

Documentation | Dashboard

MBL is a research program aims at studying the delocalization transitions with numerical methods, currently with exact diagonalization (ED) and tree tensor strong-disorder RG (tSDRG). And the support of matrix product state (MPS) is under planning.

Models

• Random-field Heisenberg model

Prerequisite

• An AWS account
• MLflow tracking server

For the purpose of better data wrangling and MLOps experience, this repo highly relies on aws data wrangler and mlflow tracking. For backend algorithms, one may refer to tnpy.

Installation

• using Docker:

  docker run --rm -it -v $(PWD)/data:/home/data tanlin2013/mbl

• using pip:

  pip install git+https://github.com/tanlin2013/mbl@main

Getting started

Run a sampler (with parallel runs supported by ray).

python scripts/sampler.py

License

© Tan Tao-Lin, 2021. Licensed under a MIT license.

Reference

D. A. Abanin, E. Altman, I. Bloch and M. Serbyn, Colloquium: Many-body localization, thermalization, and entanglement. Rev. Mod. Phys. 91, 021001 (2019).