A faster and more accurate heuristic for cyclic edit distance computation

TY - JOUR

T1 - A faster and more accurate heuristic for cyclic edit distance computation

AU - Ayad, Lorraine A.K.

AU - Barton, Carl

AU - Pissis, Solon P.

PY - 2017/3/1

Y1 - 2017/3/1

N2 - Sequence comparison is the core computation of many applications involving textual representations of data. Edit distance is the most widely used measure to quantify the similarity of two sequences. Edit distance can be defined as the minimal total cost of a sequence of edit operations to transform one sequence into the other; for a sequence x of length m and a sequence y of length n, it can be computed in time O ( m n ) . In many applications, it is common to consider sequences with circular structure: for instance, the orientation of two images or the leftmost position of two linearised circular DNA sequences may be irrelevant. To this end, an algorithm to compute the cyclic edit distance in time O ( m n log m ) was proposed (Maes, Inf. Proc. Let. 2003) and several heuristics have been proposed to speed up this computation. Recently, a new algorithm based on q-grams was proposed for circular sequence comparison (Grossi et al., Algorithms Mol Biol. 2016). We extend this algorithm for cyclic edit distance computation and show that this new heuristic is faster and more accurate than the state of the art. The aim of this letter is to give visibility to this idea in the pattern recognition community.

AB - Sequence comparison is the core computation of many applications involving textual representations of data. Edit distance is the most widely used measure to quantify the similarity of two sequences. Edit distance can be defined as the minimal total cost of a sequence of edit operations to transform one sequence into the other; for a sequence x of length m and a sequence y of length n, it can be computed in time O ( m n ) . In many applications, it is common to consider sequences with circular structure: for instance, the orientation of two images or the leftmost position of two linearised circular DNA sequences may be irrelevant. To this end, an algorithm to compute the cyclic edit distance in time O ( m n log m ) was proposed (Maes, Inf. Proc. Let. 2003) and several heuristics have been proposed to speed up this computation. Recently, a new algorithm based on q-grams was proposed for circular sequence comparison (Grossi et al., Algorithms Mol Biol. 2016). We extend this algorithm for cyclic edit distance computation and show that this new heuristic is faster and more accurate than the state of the art. The aim of this letter is to give visibility to this idea in the pattern recognition community.

KW - Cyclic edit distance

KW - circular sequences

KW - cyclic strings

KW - chain code

KW - q-gram distance

U2 - 10.1016/j.patrec.2017.01.018

DO - 10.1016/j.patrec.2017.01.018

M3 - Article

SN - 0167-8655

VL - 88

SP - 81

EP - 87

JO - PATTERN RECOGNITION LETTERS

JF - PATTERN RECOGNITION LETTERS

ER -