Learning Interpretable Temporal Properties from Positive Examples Only

We consider the problem of explaining temporal behavior black-box systems using human-interpretable models. Following recent research trends, we rely on fundamental yet interpretable models deterministic finite automata (DFAs) and linear logic (LTL_f) formulas. In contrast to most existing works for learning DFAs LTL_f formulas, from only positive examples. Our motivation is that negative examples are generally difficult observe, in particular, systems. To learn meaningful only, design algorithms conciseness language minimality as regularizers. based two approaches: a symbolic counterexample-guided one. The approach exploits an efficient encoding constraint satisfaction problem, whereas one relies generating suitable guide learning. Both approaches provide us with effective guarantees learned assess effectiveness our algorithms, evaluate them few practical case studies.