The k-mappability problem revisited

TL;DR

This paper improves algorithms for the k-mappability problem, achieving faster linear-space solutions for k=1 that work for larger alphabets and longer subwords, enhancing previous methods.

Contribution

It introduces a linear-space, optimal-time algorithm for k=1 mappability that removes the constant alphabet size restriction and extends to longer subwords.

Findings

Achieves O(n log n) time for k=1 with linear space.

Removes the constant alphabet size constraint.

Provides linear algorithms for large subword lengths.

Abstract

The $k$-mappability problem has two integers parameters $m$ and $k$. For every subword of size $m$ in a text $S$, we wish to report the number of indices in $S$ in which the word occurs with at most $k$ mismatches. The problem was lately tackled by Alzamel et al. For a text with constant alphabet $\Sigma$ and $k \in O(1)$, they present an algorithm with linear space and $O(n\log^{k+1}n)$ time. For the case in which $k = 1$ and a constant size alphabet, a faster algorithm with linear space and $O(n\log(n)\log\log(n))$ time was presented in a 2020 paper by Alzamel et al. In this work, we enhance the techniques of Alzamel et al.'s 2020 paper to obtain an algorithm with linear space and $O(n \log(n))$ time for $k = 1$. Our algorithm removes the constraint of the alphabet being of constant size. We also present linear algorithms for the case of $k=1$, $|\Sigma|\in O(1)$ and $m=\Omega(\sqrt{n})$.