A multilevel probabilistic beam search algorithm for the shortest common supersequence problem.

Abstract

The shortest common supersequence problem is a classical problem with many applications in different fields such as planning, Artificial Intelligence and especially in Bioinformatics. Due to its NP-hardness, we can not expect to efficiently solve this problem using conventional exact techniques. This paper presents a heuristic to tackle this problem based on the use at different levels of a probabilistic variant of a classical heuristic known as Beam Search. The proposed algorithm is empirically analysed and compared to current approaches in the literature. Experiments show that it provides better quality solutions in a reasonable time for medium and large instances of the problem. For very large instances, our heuristic also provides better solutions, but required execution times may increase considerably.