Online Learning of Weighted Relational Rules for Complex Event Recognition


Systems for symbolic complex event recognition detect occurrences of events in time using a set of event definitions in the form of logical rules. The Event Calculus is a temporal logic that has been used as a basis in event recognition applications, providing among others, connections to techniques for learning such rules from data. We advance the state-of-the-art by combining an existing online algorithm for learning crisp relational structure with an online method for weight learning in Markov Logic Networks (MLN). The result is an algorithm that learns complex event patterns in the form of Event Calculus theories in the MLN semantics. We evaluate our approach on a challenging real-world application for activity recognition and show that it outperforms both its crisp predecessor and competing online MLN learners in terms of predictive performance, at the price of a small increase in training time. Code related to this paper is available at

In European Conference on Machine Learning
Alexander Artikis
Alexander Artikis
Associate Professor of Artificial Intelligence