Experiments with Interval-Valued Uncertainty
Richard Tong, Lee Appelbaum
In this paper we describe the results of some recent experiments with interval representations of uncertainty. We have focussed on two basic models, one based on probability and one based on many-valued logic, and performed a series of tests to determine the resulting performance of a rule-based system for full-text information retrieval. Our results suggest that in this domain interval representations add little to our ability to capture the inherent uncertainty in the evidential reasoning process. We conjecture that concern over our ability to represent the semantics of the problem, and hence its principal reasoning structures, should dominate our concern over choices between competing formal models of uncertainty.
PDF Link: /papers/86/
AUTHOR = "Richard Tong
and Lee Appelbaum",
TITLE = "Experiments with Interval-Valued Uncertainty",
BOOKTITLE = "Uncertainty in Artificial Intelligence 2 Annual Conference on Uncertainty in Artificial Intelligence (UAI-86)",
PUBLISHER = "Elsevier Science",
ADDRESS = "Amsterdam, NL",
YEAR = "1986",
PAGES = "63--75"