Polynomial characteristic sets for DFA identification

We study the order in Grammatical Inference algorithms, and its influence on the polynomial (with respect to the data) identification of languages. This work is motivated by recent results on the polynomial convergence of data-driven grammatical inference algorithms. In this paper, we prove a sufficient condition that assures the existence of a characteristic sample whose size is polynomial with respect to the minimum DFA of the target language.