Automata learning: A categorical perspective

Abstract

Automata learning is a known technique to infer a finite state machine from a set of observations. In this paper, we revisit Angluin's original algorithm from a categorical perspective. This abstract view on the main ingredients of the algorithm lays a uniform framework to derive algorithms for other types of automata. We show a straightforward generalization to Moore and Mealy machines, which yields an algorithm already know in the literature, and we discuss generalizations to other types of automata, including weighted automata. © 2014 Springer International Publishing Switzerland.