Revista de Ingeniería
versión impresa ISSN 0121-4993
We propose a grammar inference algorithm for regular languages which saves computational cost by using two different criteria to choose states to be processed: one in the first phase of the inference process (the beginning) and another for the rest of the process. We applied experiments to observe performance of the algorithm, to learn about the best size of its first phase and to show results of its application to solve a specific problem in Bioinformatics: the cleavage site prediction problem in polyproteins encoded by viruses of the Potyviridae family.
Palabras llave : Bioinformatics; formal languages; machine learning.