Abstract:In this work, we revisit the fundamental and well-studied problem of approximate pattern matching under edit distance. Given an integer $k$, a pattern $P$ of length $m$, and a text $T$ of length $n \ge m$, the task is to find substrings of $T$ that are within edit distance $k$ from $P$. Our main result is a streaming algorithm that solves the problem in $\tilde{O}(k^5)$ space and $\tilde{O}(k^8)$ amortised time per character of the text, providing answers correct with high probability. (Hereafter, $\tilde{O}(\cdot)$ hides a $\mathrm{poly}(\log n)$ factor.) This answers a decade-old question: since the discovery of a $\mathrm{poly}(k\log n)$-space streaming algorithm for pattern matching under Hamming distance by Porat and Porat [FOCS 2009], the existence of an analogous result for edit distance remained open. Up to this work, no $\mathrm{poly}(k\log n)$-space algorithm was known even in the simpler semi-streaming model, where $T$ comes as a stream but $P$ is available for read-only access. In this model, we give a deterministic algorithm that achieves slightly better complexity. In order to develop the fully streaming algorithm, we introduce a new edit distance sketch parametrised by integers $n\ge k$. For any string of length at most $n$, the sketch is of size $\tilde{O}(k^2)$ and it can be computed with an $\tilde{O}(k^2)$-space streaming algorithm. Given the sketches of two strings, in $\tilde{O}(k^3)$ time we can compute their edit distance or certify that it is larger than $k$. This result improves upon $\tilde{O}(k^8)$-size sketches of Belazzougui and Zhu [FOCS 2016] and very recent $\tilde{O}(k^3)$-size sketches of Jin, Nelson, and Wu [STACS 2021].

What problem does this paper attempt to address?

The problem that this paper attempts to solve is approximate pattern matching with edit - distance constraints in the streaming data model. Specifically, given an integer $k$, a pattern $P$ of length $m$, and a text $T$ of length $n$ ($n\geq m$), the task is to find substrings in the text $T$ whose edit distance from the pattern $P$ is no more than $k$. The edit distance refers to the minimum number of operations (insertion, deletion, or substitution of characters) required to make two strings identical. ### Main contributions of the paper 1. **Streaming algorithm**: - A streaming algorithm is proposed, which solves the problem with $\tilde{O}(k^{5})$ space and $\tilde{O}(k^{8})$ amortized time complexity per character, and the accuracy rate of the answer is very high. Here, $\tilde{O}(\cdot)$ represents hiding the $\text{poly}(\log n)$ factor. - This result answers a decade - long question: Since Porat and Porat [FOCS 2009] discovered the multi - logarithmic - space streaming algorithm based on the Hamming distance, whether there exists a similar edit - distance streaming algorithm has been an open question. 2. **Deterministic algorithm in the semi - streaming model**: - In the semi - streaming model, the text arrives in the form of a stream, but the pattern can be accessed in read - only mode. In this model, a deterministic algorithm is proposed, with a space complexity of $\tilde{O}(k^{5})$ and an amortized time complexity of $\tilde{O}(k^{6})$ per character. 3. **New edit - distance sketch**: - A new edit - distance sketch is designed to retrieve the edit distances (up to $k$) between strings of length no more than $n$. The size of this sketch is $\tilde{O}(k^{2})$, and it can calculate the edit distance between two strings or confirm that it is greater than $k$ in $\tilde{O}(k^{2})$ space and $\tilde{O}(k^{3})$ time. - This result improves the $\tilde{O}(k^{8})$ - sized sketch of Belazzougui and Zhang [FOCS 2016] and the $\tilde{O}(k^{3})$ - sized sketch of Jin, Nelson, and Wu [STACS 2021]. ### Technical contributions 1. **Greedy encoding**: - A new space - efficient deterministic encoding method, called Greedy Encoding, is introduced to encode all alignments of two strings with costs no more than $k$. - Through this encoding, substrings in the text close to the pattern can be compressed. Specifically, for strings of length no more than $n$, the encoding occupies $\tilde{O}(k^{2})$ space. 2. **Edit - distance sketch**: - A new edit - distance sketch with the parameter of an integer $n\geq k$ is designed. For any string of length no more than $n$, the size of the sketch is $\tilde{O}(k^{2})$, and it can be calculated by a streaming algorithm with $\tilde{O}(k^{2})$ space. - Given the sketches of two strings, their edit distance can be calculated or it can be confirmed that it is greater than $k$ in $\tilde{O}(k^{3})$ time. ### Conclusion By introducing new encoding methods and edit - distance sketches, this paper successfully solves the approximate pattern - matching problem with edit - distance constraints in the streaming data model, fills the gap in this field, and provides new tools and methods for future related research.

Small space and streaming pattern matching with k edits

Streaming $k$-edit approximate pattern matching via string decomposition

$L_p$ Pattern Matching in a Stream

On the Communication Complexity of Approximate Pattern Matching

The Streaming k-Mismatch Problem: Tradeoffs between Space and Total Time

Effective and Efficient Shape-Based Pattern Detection over Streaming Time Series

Approximate Circular Pattern Matching under Edit Distance

Faster Sublinear-Time Edit Distance

Edit Distance: Sketching, Streaming and Document Exchange

Near-Optimal-Time Quantum Algorithms for Approximate Pattern Matching

Almost Linear Size Edit Distance Sketch

Longest Alignment with Edits in Data Streams

Streaming Algorithms for Graph k-Matching with Optimal or Near-Optimal Update Time

Faster two-dimensional pattern matching with $k$ mismatches

Faster Algorithms for Text-to-Pattern Hamming Distances

An Improved Sketching Algorithm for Edit Distance

Elastic-Degenerate String Matching with 1 Error

An Improved Algorithm for The $k$-Dyck Edit Distance Problem

Bounded Edit Distance: Optimal Static and Dynamic Algorithms for Small Integer Weights

Text Searching Allowing for Non-Overlapping Adjacent Unbalanced Translocations

Pattern Matching with Mismatches and Wildcards