Abstract
\(\mathcal{R}\mathcal{E}\mathcal{S}\) is a system for reasoning about evidential support relationships between statements[1, 2]. In \(\mathcal{R}\mathcal{E}\mathcal{S}\), the preferences of these supports are represented symbolically, by directly comparing them, instead of by numerical degrees. Z+ is a formalism for reasoning with variable-strength defaults[5] which provides a mechanism to compute a minimum admissible ranking for models (subject to the consistency condition) from the given integer strengths of defaults.
In this paper, we combine the two systems. We show that the same consistency condition of Z + can be applied to \(\mathcal{R}\mathcal{E}\mathcal{S}\) even though the preferences of rules are represented as a relation in \(\mathcal{R}\mathcal{E}\mathcal{S}\). A similar procedure is devised which can produce the admissible relative strengths (a relation) and can produce the relation on models with respect to the strengths of the rules they violate. A consequence relation is defined and a procedure to answer queries concerning it is devised. The resulting system, also called \(\mathcal{R}\mathcal{E}\mathcal{S}\), is then compared to Z +. We show that, while \(\mathcal{R}\mathcal{E}\mathcal{S}\) is very similar to Z + and displays comparable reasoning processes most of the time, they are not the same and \(\mathcal{R}\mathcal{E}\mathcal{S}\) is more in agreement with common sense in some situations. Comparing \(\mathcal{R}\mathcal{E}\mathcal{S}\) to the stratified ranking system [6] shows that \(\mathcal{R}\mathcal{E}\mathcal{S}\), as presented, also shares some limitations with Z +
Preview
Unable to display preview. Download preview PDF.
References
Z. An, D. Bell, and J. Hughes. \(\mathcal{R}\mathcal{E}\mathcal{S}\)—a relation based method for evidential reasoning. In Proc. 8th Conf. on Uncertainty in AI, pages 1–8, 1992.
Z. An, D. Bell, and J. Hughes. Relation based evidential reasoning. Int. J. on Approximate. Reasoning, accepted, in preparation.
P. P. Bonissone et al. Uncertainty and incompleteness: Breaking the symmetry of defeasible reasoning. In M.Henrion, R.D.Shachter, L.N.Kanal, and J.F.Lemmer, editors, Uncertainty in Artificial Intelligence, volume 5, pages 67–85. 1990.
C. Boutilier. What is a default priority? In Proc. CCAI 1992, pages 140–147, 1992.
M. Goldszmidt and J. Pearl. System-Z+: A formalism for reasoning with variable-strength defaults. In Proc. AAAI-91, 1991.
M. Goldszmidt and J. Pearl. Rank-based systems: A simple approach to belief revision, belief update, and reasoning about evidence and actions. In Proc. Conf. Knowledge Representation, 1992.
B. Grosof. Generalizing prioritization. In J. Allen, J. Fikes, and E. Sandewall, editors, Principles of Knowledge. Representation: Proc. of the 2nd. Int. Conf. Morgan-Kauffmann, 1991.
J. Y. Halpern and M. O. Rabin. A logic to reason about likelihood. Artificial Intelligence, 32:379–405, 1987.
S. Kraus, D. Lehmann, and M. Magidor. Nonmonotonic reasoning, preferential models and cumulative logics. Artificial Intelligence, 44:167–207, 1990.
H. Kyburg Jr. The reference class. Philosophy of Science, pages 374–97, 1983.
V. Lifschitz. Circumscriptive theories: a logic-based framework for knowledge representation. Journal of Philosophical Logic, 17:391–441, 1988.
J. Pearl. Probabilistic Reasoning in Intelligent Systems: Networks for Plausible Inference. Kaufman, 1988.
G. Shafer. A Mathematical Theory of Evidence. Princeton University Press, 1976.
Author information
Authors and Affiliations
Editor information
Rights and permissions
Copyright information
© 1993 Springer-Verlag Berlin Heidelberg
About this paper
Cite this paper
An, Z., McLeish, M. (1993). \(\mathcal{R}\mathcal{E}\mathcal{S}\): A formalism for reasoning with relative-strength defaults. In: Clarke, M., Kruse, R., Moral, S. (eds) Symbolic and Quantitative Approaches to Reasoning and Uncertainty. ECSQARU 1993. Lecture Notes in Computer Science, vol 747. Springer, Berlin, Heidelberg. https://doi.org/10.1007/BFb0028175
Download citation
DOI: https://doi.org/10.1007/BFb0028175
Published:
Publisher Name: Springer, Berlin, Heidelberg
Print ISBN: 978-3-540-57395-1
Online ISBN: 978-3-540-48130-0
eBook Packages: Springer Book Archive