Multidimensional decoding networks for trapping set analysis

Abstract

We present a novel multidimensional network model as a means to analyze decoder failure and characterize trapping sets of graph-based codes. We identify a special class of these decoding networks, which we call transitive networks, and show how they may be used to identify trapping sets and inducing sets. Many codes have transitive decoding network representations. We conclude by investigating the decoding networks of codes arising from product, half-product, and protograph code constructions.