Correctness and termination of Tomte components for active register automata learning

In this thesis, we will analyze some of the components of the learning environment of the learning tool Tomte. This tool adds components to the learning environment, which enables it to learn a large class of extended finite state machines, including register automata with fresh output values. We will describe the Determinizer, a component that transforms values of register automata with non-deterministic behavior and the Lookahead Oracle, a component that enriches traces with extra information during the learning process. We will show the correctness of the Determinizer component. Furthermore, we will give an upper bound for lookahead traces that act as a witness for memorable values, thus showing that the Lookahead Oracle can find all memorable values by running finitely many lookahead traces.