Learning Nominal Regular Languages with Binders
2020-01-17T16:13:38Z (GMT) by
The theories of formal language and automata are fundamental in several areas of computer science. During decades of development, these theories have been stretched out and reach many branches and application contexts, ranging from lexical analysis, natural language processing, model checking, and system design. Recently, the applications of machine learning are spreading out rapidly. One interesting application, learning automata, gains sustained attention crossing the disciplines.
This dissertation investigates learning nominal automata, an extension of classical automata to alphabets featuring names. This class of automata capture nominal regular languages; analogously to the classical language theory, nominal automata have been shown to characterise nominal regular expressions with binders. These formalisms are amenable to abtract modelling resource-aware computations.
We propose nL*, a learning algorithm on nominal regular languages with binders. Our algorithm generalises Angluin's algorithm L* with respect to nominal regular languages with binders. We show the correctness of nL* and study its theoretical complexity.
We also develop a implementation of nL* that we use to experimentally analyse different strategies for providing counterexamples to the learner. These strategies are designed on the base of the rich algebraic structure provided by our nominal setting. Further, we evaluate the behaviours of the process and its efficiency according to
the different strategies with a set of experiments.