Inferring Automata and its Application to Compositional Verification WPEII : Critical Review Report

Given an oracle which can answer membership and equivalence queries for an unknown regular language, we present algorithms for learning the regular set perfectly. For the case where the oracle cannot reset the black box automaton to a fixed initial state, we present efficient learning algorithms using homing sequences. Algorithms have been described for two different representations of the finite automaton. In the special case where the automaton is a permutation automaton, we present algorithms which can deduce the unknown regular set without the use of equivalence queries. We also describe a modified learning algorithm which can be used to learn context-free grammars from their structural descriptions. Application of these learning techniques for compositional verification of symbolic modules has also been discussed. We also describe improvements to the learning algorithm for context-free languages under certain restrictions on the result of equivalence queries.