**BSR lecture**: Tuesday July 3rd, 14:00-15:30

**Title**: Learning State Machines using Flexfringe

**Abstract**: Flexfringe is a recent algorithm for learning finite state machines based on the classic red-blue fringe state merging technique. It can be used to learn different types of state machines such as finite state automata, probabilistic automata, Mealy machines, regression automata, and real-time automata. The only real constraints on the learned model are that it has to be deterministic, and that a Markov property or Myhill-Nerode congruence holds. The semantics of states and how this is implemented is flexible. In this talk, I will explain the inner workings of the state merging methods and different ways of using the state flexibility. There exist many different applications of state machine learning and Flexfringe contains many parameters and different learning approaches that allows you to tune and significantly modify the algorithm output. I will demonstrate ways to set these parameters and how to choose a successful learning approach for different kinds of software systems.

**Short bio**: **Sicco Verwer** is an assistant professor of cyber security and machine learning at Delft University of Technology.