Knowledge Discovery in Biosequences Using Sort Regular Patterns

This paper considers knowledge discovery by sort regular patterns, which are strings over sort letters representing nite sets of basic letters. We devise a learning algorithm for the class based on the minimal multiple generalization technique, and evaluate the method by experiments on biosequences from GenBank database. The experiments show that relatively a simple sort pattern can represent a complex motif in biosequences, and the learning algorithm works well in noisy examples.