Lisette Sanchez: Learning software behavior through active automata learning with data

Event Details

The topic of the thesis is find out whether the LS* learning algorithm, that can learn register automata with abstract data parameters from actual software is practically applicable in an industrial context. The algorithm can for instance learn a queue with limited size that stores arbitrary natural numbers.

The conclusion is that indeed practical software can be learned, outperforming the learning of automata with concrete data. Scale remains a problem, though.