Rick Smetsers

Minimal Separating Sequences for All Pairs of States

Finding minimal separating sequences for all pairs of inequivalent states in a finite state machine is a classic problem in automata theory. Sets of minimal separating sequences, for instance, play a central role in many conformance testing methods. Moore has already outlined a partition refinement algorithm that constructs such a set of sequences in \(\mathcal {O}(mn)\) time, where m is the number of transitions and n is the number of states. In this paper, we present an improved algorithm based on the minimization algorithm of Hopcroft that runs in \(\mathcal {O}(m \log n)\) time. The efficiency of our algorithm is empirically verified and compared to the traditional algorithm.

Model Learning as a Satisfiability Modulo Theories Problem

We explore an approach to model learning that is based on using satisfiability modulo theories (SMT) solvers. To that end, we explain how DFAs, Mealy machines and register automata, and observations of their behavior can be encoded as logic formulas. An SMT solver is then tasked with finding an assignment for such a formula, from which we can extract an automaton of minimal size. We provide an implementation of this approach which we use to conduct experiments on a series of benchmarks. These experiments address both the scalability of the approach and its performance relative to existing active learning tools.

Enhancing Automata Learning by Log-Based Metrics

We study a general class of distance metrics for deterministic Mealy machines. The metrics are induced by weight functions that specify the relative importance of input sequences. By choosing an appropriate weight function we may fine-tune a metric so that it captures some intuitive notion of quality. In particular, we present a metric that is based on the minimal number of inputs that must be provided to obtain a counterexample, starting from states that can be reached by a given set of logs. For any weight function, we may boost the performance of existing model learning algorithms by introducing an extra component, which we call the Comparator. Preliminary experiments show that use of the Comparator yields a significant reduction of the number of inputs required to learn correct models, compared to current state-of-the-art algorithms. In existing automata learning algorithms, the quality of subsequent hypotheses may decrease. Generalising a result of Smetsers eti¾?al., we show that the quality of hypotheses that are generated by the Comparator never decreases.