Learning ω-regular languages from queries and counter-examples (a preliminary report)

Abstract

In this paper we discuss problems concerning the learnability of ω-regular languages from examples via the synthesis of their corresponding finite-state acceptors. The main technical contribution of this paper is a polynomial algortihm that learns a sub-class of the ω-regular languages from queries and counter-examples. A similar setting has been suggested in [Ang87] for learning (finitary) regular languages. The algorithm given there produces the minimal automaton accepting the desired language using polynomially many (in the number of states) queries and computation steps. Our algorithm tackles some special problems that arise when adapting this framework to the infinitary case.