CIKM '97 Proceedings of the sixth international conference on Information and knowledge management
Probabilistic logic programming with conditional constraints
ACM Transactions on Computational Logic (TOCL)
Column Generation Methods for Probabilistic Logic
Proceedings of the 1st Integer Programming and Combinatorial Optimization Conference
Expressive probabilistic description logics
Artificial Intelligence
ISWC '08 Proceedings of the 7th International Conference on The Semantic Web
Laconic and Precise Justifications in OWL
ISWC '08 Proceedings of the 7th International Conference on The Semantic Web
Probabilistic deduction with conditional constraints over basic events
Journal of Artificial Intelligence Research
Russian doll search for solving constraint optimization problems
AAAI'96 Proceedings of the thirteenth national conference on Artificial intelligence - Volume 1
Trust-based probabilistic query answering
WISE'11 Proceedings of the 12th international conference on Web information system engineering
Description logics and fuzzy probability
IJCAI'11 Proceedings of the Twenty-Second international joint conference on Artificial Intelligence - Volume Volume Two
Hi-index | 0.00 |
This paper presents an approach aimed at improving scalability reasoning in the probabilistic description logic P-$\mathcal{SHIQ}$(D). The satisfiability problem is currently intractable as the existing algorithm generates exponentially large linear optimization systems. To cope with this, we employ advanced optimization techniques that do not require the systems to be fully generated before starting to solve them. One such technique, namely, column generation, can be applied by exploiting the structure of the reasoning problems in P-$\mathcal{SHIQ}$(D) analogously to how it has been used for solving such large optimization problems as constrained shortest path, cutting stock and many others. We present the formulation of the linear systems for the satisfiability problems in the column generation form, describe methods for solving them, discuss their difference from the known column generation approaches to propositional probabilistic logic, and finally show preliminary experimental results that demonstrate a useful improvement of scalability.