A prototype theory interpretation of the label semantics framework is proposed as a possible model of imprecise descriptions of real numbers. It is shown that within this framework conditioning given imprecise descriptions of a real variable naturally results in imprecise probabilities. An inference method is proposed from data in the form of a set of imprecise descriptions, which naturally suggests an algorithm for estimating lower and upper probabilities given imprecise data values.
Keywords. Label Semantics, Prototype Theory, Random Sets, Lower and Upper Distributions, Second Order Distributions
The paper is availabe in the following formats:
Plenary talk : Press here to get the file of the presentation.
Dept. Engineering Mathematics,
University of Bristol,
Bristol, BS8 1TR
Dept. Maths, Stats and Compt.
University of Cantabria
College of Computer Science