Veröffentlichung

Titel:

Towards a More Efficient Computation of Weighted Conditional Impacts for Relational Probabilistic Knowledge Bases under Maximum Entropy Semantics

AutorInnen:

Marc Finthammer
Christoph Beierle

Kategorie:

Konferenzbandbeiträge

erschienen in:

Proceedings of the 38th Annual German Conference on Advances in Artificial Intelligence (KI-2015), pp. 72--86 (2015)

Abstract:

While the complexity of the optimization problem to be solved when computing the Maximum Entropy distribution P^*_R of a knowledge base R grows dramatically when moving to the relational case, it has been shown that having the weighted conditional impacts (WCI) of R available, P^*_R can be computed much faster. Computing WCI in a straightforward manner readily gets infeasible due to the size of the set Ω of possible worlds. In this paper, we propose a new approach for computing the WCI without considering the worlds in Ω at all. We introduce the notion of sat-pairs and show how to determine the set CSP of all possible combinations of sat-pairs by employing combinatorial means. Using CSP instead of Ω for computing the WCI is a significant performance gain since CSP is typically much smaller than Ω. For a start, we focus on simple knowledge bases consisting of a single conditional. First evaluation results of an implemented algorithm illustrate the benefits of our approach.

Download:

Springer

BibTeX-Eintrag:

@InProceedings{FinthammerBeierle2015KI, author = {Marc Finthammer and Christoph Beierle}, title = {Towards a More Efficient Computation of Weighted Conditional Impacts for Relational Probabilistic Knowledge Bases under Maximum Entropy Semantics}, booktitle = {Proceedings of the 38th Annual German Conference on Advances in Artificial Intelligence ({KI}-2015)}, year = {2015}, pages = {72--86}, url = {http://dx.doi.org/10.1007/978-3-319-24489-1\_6}, }