Efficient Approximation Algorithms for String Kernel Based Sequence Classification

TL;DR

This paper introduces efficient approximation algorithms for string kernel-based sequence classification, enabling the use of larger parameters and improving accuracy across various data types, with theoretical guarantees and empirical validation.

Contribution

The authors develop novel algorithms that accurately estimate sequence similarity scores, overcoming computational limitations and solving an open combinatorial problem for scalable sequence classification.

Findings

Achieves high-quality approximation with theoretical bounds.

Enables use of larger k and m parameters for better accuracy.

Demonstrates effectiveness on biological and music datasets.

Abstract

Sequence classification algorithms, such as SVM, require a definition of distance (similarity) measure between two sequences. A commonly used notion of similarity is the number of matches between $k$-mers ($k$-length subsequences) in the two sequences. Extending this definition, by considering two $k$-mers to match if their distance is at most $m$, yields better classification performance. This, however, makes the problem computationally much more complex. Known algorithms to compute this similarity have computational complexity that render them applicable only for small values of $k$ and $m$. In this work, we develop novel techniques to efficiently and accurately estimate the pairwise similarity score, which enables us to use much larger values of $k$ and $m$, and get higher predictive accuracy. This opens up a broad avenue of applying this classification approach to audio, images, and text sequences. Our algorithm achieves excellent approximation performance with theoretical guarantees. In the process we solve an open combinatorial problem, which was posed as a major hindrance to the scalability of existing solutions. We give analytical bounds on quality and runtime of our algorithm and report its empirical performance on real world biological and music sequences datasets.