Reputational Genetic Model for Regular Inference

Abstract

The regular inference is one of the main problems of the formal language theory, which is to synthesize a finite-state automaton corresponding to some unknown regular language represented with a list of positive and negative examples. In this paper, we propose a new algorithm for regular inference along with special measures for evaluating quality of elements of automaton we call reputation. The algorithm belongs to genetic algorithms family and transforms candidate automatons based on the reputation of its elements. We prove effectiveness of our model by experiments on pregenerated datasets.