AAAI Publications, Twenty-Third International FLAIRS Conference

Font Size: 
A System for Relational Probabilistic Reasoning on Maximum Entropy
Matthias Thimm, Marc Finthammer, Sebastian Loh, Gabriele Kern-Isberner, Christoph Beierle

Last modified: 2010-05-06

Abstract


Comparisons of different approaches to statistical relational learning are difficult due to the variety of the available concepts and due to the absense of a common interface. The main objective of the KReator toolbox introduced here is to provide a common methodology for modelling, learning, and inference in a relational probabilistic framework. As a second major contribution of this paper, we present the RME approach to relational probabilistic reasoning which applies the principle of maximum entropy to groundings of a relational knowledge base and which is also supported by KReator.

Full Text: PDF