Contextual probability
DOI:
https://doi.org/10.26636/jtit.2003.3.186Keywords:
mathematical foundations, , knowledge representation, machine learning, uncertainty, data miningAbstract
In this paper we present a new probability function G that generalizes the classical probability function. A mass function is an assignment of basic probability to some context (events, propositions). It represents the strength of support for some contexts in a domain. A context is a subset of the basic elements of interest in a domain - the frame of discernment. It is a medium to carry the "probabilistic" knowledge about a domain. The G function is defined in terms of a mass function under various contexts. G is shown to be a probability function satisfying the axioms of probability. Therefore G has all the properties attributed to a probability function. If the mass function is obtained from probability function by normalization, then G is shown to be a linear function of probability distribution and a linear function of probability. With this relationship we can estimate probability distribution from probabilistic knowledge carried in some contexts without any model assumption.
Downloads
References
[1] W. Feller, An Introduction to Probability Theory and its Applications. Wiley, 1968.
View in Google Scholar
[2] J. W. Guan and D. A. Bell, "Generalization of the Dempster-Shafer theory", in Proc. IJCAI-93, 1993, pp. 592-597.
View in Google Scholar
[3] D. Hand, H. Mannila, and P. Smyth, Principles of Data Mining. The MIT Press, 2001.
View in Google Scholar
[4] E. T. Jaynes, "Probability theory: the logic of science", http://bayes.wustl.edu
View in Google Scholar
[5] B. D. Ripley, Pattern Recognition and Neural Networks. Cambridge University Press, 1996.
View in Google Scholar
[6] G. Shafer, A Mathematical Theory of Evidence. Princeton, New Jersey: Princeton University Press, 1976.
View in Google Scholar
Downloads
Published
Issue
Section
License
Copyright (c) 2003 Journal of Telecommunications and Information Technology

This work is licensed under a Creative Commons Attribution 4.0 International License.