Improved Singleton bound on insertion-deletion codes and optimal constructions

TL;DR

This paper improves the upper bound on the minimum insertion-deletion distance for linear codes, generalizes previous results, and constructs explicit optimal Reed-Solomon codes meeting the new bound.

Contribution

It establishes a tighter bound on insdel distance for all linear codes with n>k>1 and provides explicit constructions of optimal two-dimensional Reed-Solomon codes.

Findings

The minimum insdel distance of any [n,k] linear code over f_q is at most 2n-2k for n>k>1.

A sufficient condition for a two-dimensional Reed-Solomon code to have insdel distance exactly 2n-4.

Explicit construction of an infinite family of optimal two-dimensional Reed-Solomon codes.

Abstract

Insertion-deletion codes (insdel codes for short) play an important role in synchronization error correction. The higher the minimum insdel distance, the more insdel errors the code can correct. Haeupler and Shahrasbi established the Singleton bound for insdel codes: the minimum insdel distance of any $[n,k]$ linear code over $\mathbb{F}_q$ satisfies $d\leq2n-2k+2.$ There have been some constructions of insdel codes through Reed-Solomon codes with high capabilities, but none has come close to this bound. Recently, Do Duc {\it et al.} showed that the minimum insdel distance of any $[n,k]$ Reed-Solomon code is no more than $2n-2k$ if $q$ is large enough compared to the code length $n$; optimal codes that meet the new bound were also constructed explicitly. The contribution of this paper is twofold. We first show that the minimum insdel distance of any $[n,k]$ linear code over $\mathbb{F}_q$ satisfies $d\leq2n-2k$ if $n>k>1$. This result improves and generalizes the previously known results in the literature. We then give a sufficient condition under which the minimum insdel distance of a two-dimensional Reed-Solomon code of length $n$ over $\mathbb{F}_q$ is exactly equal to $2n-4$. As a consequence, we show that the sufficient condition is not hard to achieve; we explicitly construct an infinite family of optimal two-dimensional Reed-Somolom codes meeting the bound.