A Python package using mostly JAX to implement the time-evolving block decimation algorithm in a differentiable way along with some data generation tools to synthesize Hamiltonian learning problems.
This package was used to generate the numerical results in our paper on Scalably learning quantum many-body Hamiltonians from dynamical data.
The default type for complex-valued arrays is double precision, i.e. jax.numpy.complex128.
To work with single precision, set the environment variable TEBD_COMPLEX_TYPE=complex64 before importing this package.
Install the package via pip install -e /path/to/differentiable-tebd.
The only dependency that is not installed automatically is jaxlib.
You need to install it manually as this differs depending on the hardware you want to use (installation guide).
Be sure to checkout the relevant or latest version tag (e.g. git checkout 0.0.0) to be able to reproduce results easily.
For a brief demo for how to use this package please have a look at our repository for the paper.
If you use (parts of) this package, please cite our paper.
@misc{wilde_scalably_2022,
doi = {10.48550/ARXIV.2209.14328},
url = {https://arxiv.org/abs/2209.14328},
author = {Wilde, Frederik and Kshetrimayum, Augustine and Roth, Ingo and Hangleiter, Dominik and Sweke, Ryan and Eisert, Jens},
title = {Scalably learning quantum many-body Hamiltonians from dynamical data},
publisher = {arXiv},
year = {2022},
copyright = {Creative Commons Attribution 4.0 International}
}