Edge-Based Anisotropic Decoding for Generalized Bicycle Codes

TL;DR

This paper introduces an edge-based anisotropic decoding method for generalized bicycle quantum codes, improving iterative decoding performance by exploiting graph symmetries and breaking degeneracies.

Contribution

It provides a graph-theoretic framework for understanding degeneracy in GB codes and proposes an edge-coloring approach to enhance decoding effectiveness.

Findings

Edge-coloring eliminates automorphisms in low-weight subgraphs.

Edge anisotropic decoding outperforms isotropic methods in experiments.

Improved decoding performance achieved in fewer iterations.

Abstract

Quantum low-density parity-check (QLDPC) codes provide non vanishing rates, distance scaling with the blocklength of the code, and facilitate fast iterative decoding because of their sparsity. However, in practice iterative decoding fails to exploit the distance of the code, because it cannot resolve the symmetries imposed by degeneracy. In this work, we provide a graph theoretic characterization of degeneracy for the family of generalized bicycle (GB) codes. This viewpoint shows that harmful degenerate error patterns persist whenever they remain related by automorphisms preserved by the decoder. Motivated by symmetry breaking via graph coloring, we compare three coloring approaches: no coloring, block-coloring, and edge-coloring. For GB codes, we show that edge-coloring can eliminate all automorphisms in low-weight stabilizer-induced subgraphs. We practically realize the coloring schemes as isotropic, block- anisotropic and edge-anisotropic min-sum (MS) decoding. Experimental results show that edge anisotropic min-sum decoding obtains improved performance over isotropic and block anisotropic decoding for several GB codes in a small number of iterations.