Upper Bounds on the Minimum Distance of Structured LDPC Codes

2025-01-31

AI Summary

This work investigates the asymptotic minimum distance upper bound for binary LDPC codes with a [C|M] structure, where C is a column-weight-2 circulant matrix and M is a matrix of column weight r ≥ 3 and row weight at least 1—aiming to characterize the fundamental limits of their error-correction capability. Such structured codes admit linear-time encoding while maintaining strong decoding performance. We present the first improvement to the long-standing asymptotic minimum distance upper bound, tightening it from O(n^{(r−1)/r}) to O(n^{(r−2)/(r−1)+ε}) for any ε > 0. Our approach integrates combinatorial construction analysis, algebraic properties of circulant matrices, low-weight codeword enumeration via Tanner graph expansion, and asymptotic graph-theoretic modeling. The resulting bound constitutes the best-known upper bound to date, establishing a critical theoretical constraint and performance benchmark for the design of efficiently encodable LDPC codes.

Technology Category

• Machine Learning: Structured Learning
• Constraint Satisfaction and Optimization: Distributed CSP/Optimization
• Cognitive Modeling & Cognitive Systems: Neural Spike Coding

Application Category

• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs
• Web Mining and Content Analysis: Bridging structured and unstructured data
• Economics, Online Markets and Human Computation: Incentives in network design for Web infrastructures and ecosystems

Abstract

We investigate the minimum distance of structured binary Low-Density Parity-Check (LDPC) codes whose parity-check matrices are of the form $[mathbf{C} vert mathbf{M}]$ where $mathbf{C}$ is circulant and of column weight $2$, and $mathbf{M}$ has fixed column weight $r geq 3$ and row weight at least $1$. These codes are of interest because they are LDPC codes which come with a natural linear-time encoding algorithm. We show that the minimum distance of these codes is in $O(n^{frac{r-2}{r-1} + epsilon})$, where $n$ is the code length and $epsilon>0$ is arbitrarily small. This improves the previously known upper bound in $O(n^{frac{r-1}{r}})$ on the minimum distance of such codes.

Problem

• Structured Binary LDPC Codes
• Minimum Distance Bound
• Error Resistance

Innovation

• Structured Binary LDPC Codes
• Minimum Distance
• Error Resistance