Unit:
Department of Informatics and TelecommunicationsΠληροφορική
Author:
NEAMONITIS EVANGELOS
Supervisors info:
Παναγιώτης Σταματόπουλος, Επίκουρος Καθηγητής, ΕΚΠΑ
Νικόλαος Κατζούρης, Βοηθός Ερευνητή, ΕΚΕΦΕ Δημόκριτος
Original Title:
Distributed Online Learning of Probabilistic LogicalTheories for Complex Event Recognition
Translated title:
Distributed Online Learning of Probabilistic LogicalTheories for Complex Event Recognition
Summary:
Complex Event Recognition (CER) systems detect occurrences of complex events (e.g meeting, moving, dangerous driving) in a streaming time-stamped input of simple events using known event patterns. Logic-based approaches that are able to learn Event Calculus theories are of particular interest in CER, as they can effectively deal with uncertainty and noise in data streams, thus being a robust alternative to the costly process of manually crafting event pattern theories. In this context WOLED has been presented. WOLED is a system that is based on Answer Set Programming (ASP). In recent years, the amount of data produced has seen an unprecedented increase. CER systems, ought to be able to cope with this and scale to the need of a given application. We focus on ways to tackle this rising problem by attempting to perceive WOLED in a distributed learning scenario with multiple learners. In this study, we compare and evaluate different possible communication protocols for sharing learned complex event patterns between learners.
Main subject category:
Technology - Computer science
Keywords:
Inductive Logic Programming ,Event Calculus, Distributed Systems, Complex Event Recognition, Geometric Monitoring
