Towards ILP-based LTLf passive learning

Abstract Inferring linear temporal logic over finite traces ($\text{LTL}_{\text{f}}$) formulas from a set of example traces, known as passive learning, presents significant challenges due to its combinatorial nature. In this paper, we introduce a novel approach to $\text{LTL}_{\text{f}}$ passive learning based on inductive logic programming (ILP), leveraging the inductive learning of answer set programs framework. Our ILP-based method effectively exploits the set of example traces to guide the learning process, and experimental results demonstrate that it o ffers a more efficient solution compared to traditional techniques based on propositional satisfiability.