Online probabilistic theory revision from examples : a proPPR approach

Abstract

Handling relational data streams has become a crucial task, given the availability of pervasive sensors and Internet-produced content, such as social networks and knowledge graphs. In a relational environment, this is a particularly challenging task, since one cannot assure that the streams of examples are independent along the iterations. Thus, most relational machine learning methods are still designed to learn only from closed batches of data, not considering the models acquired in previous iterations of incoming examples. In this work, we propose OSLR, an online relational learning algorithm that can handle continuous, open-ended streams of relational examples as they arrive. We employ techniques from theory revision to take advantage of the already acquired knowledge as a starting point, find where it should be modified to cope with the new examples, and automatically update it. We rely on the Hoeding’s bound statistical theory to decide if the model must in fact be updated accordingly to the new examples. Our system is built upon ProPPR statistical relational language to describe the induced models, aiming at contemplating the uncertainty inherent to real data. Experimental results in entity co-reference and social networks datasets show the potential of the proposed approach compared to other relational learners