autdec 0.1.0

Automorphism Ensemble Decoder for Quantum Stabilizer Codes

AutDEC: Quantum Automorphism Ensemble Decoder

Python package for implementation of the Quantum Automorphism Ensemble Decoder, "AutDEC".

AutDEC is designed for parallel ensemble decoding of quantum error-correcting codes with large automorphism groups. It works by creating an ensemble of automorphism guided BP, BP+OSD or BP+LSD decoders. For Bivariate Bicycle Codes under circuit-level noise, using only a BP-based ensemble, we approach and match BP+OSD0 performance without postprocessing. On the other hand, using a BP+OSD0-based ensemble, we achieve similar decoding performance to BP+OSD10. It is inspired by the work of M. Geiselhart et al for classical codes [1].

Note: AutDEC is currently a work in progress and a serial implementation is provided with parallelization in future plans.

Installation

PyPI: pip install autdec

Source code: Download this repository and run pip install .

Overview

Below we have an overview of the AutDEC online decoding phase:

The effectiveness of BP is limited by the presence of short-cycles on the Tanner graph. Automorphisms allow AutDEC to bypass this problem as seen below:

We provide functions for finding the graph automorphism group of arbitrary codes.

Results

In Code Capacity we provide a notebook and simulation results for [[15,1,3]] punctured Reed-Muller code in depolarising code capacity noise model.

For Bivariate Bicycle Codes we provide a notebook and simulation results under circuit-level noise simulations. Currently, simulations can take a long time for large codes because of the serial implementation.

Citation

@misc{Sayginel_AutDEC_2025,
author = {Sayginel, Hasan and Koutsioumpas, Stergios},
license = {MIT},
month = mar,
title = {{AutDEC}},
url = {https://github.com/hsayginel/autdec},
version = {1.0.0},
year = {2025}
}

References

[1] M. Geiselhart, A. Elkelesh, M. Ebada, S. Cammerer and S. t. Brink, "Automorphism Ensemble Decoding of Reed–Muller Codes", in IEEE Transactions on Communications, vol. 69, no. 10, pp. 6424-6438, Oct. 2021, doi: 10.1109/TCOMM.2021.3098798.