On Open Problems Connected with Application of the Iterative Proportional Fitting Procedure to Belief Functions

0395348 - ÚTIA 2014 RIV FR eng C - Conference Paper (international conference)
Kratochvíl, Václav - Jiroušek, Radim
On Open Problems Connected with Application of the Iterative Proportional Fitting Procedure to Belief Functions.
Proceedings of the Eighth International Symposium on Imprecise Probability: Theories adn Applications. Compiegne: Society for Imprecise Probability: Theories and Applications, 2013, s. 149-158. ISBN 978-2-913923-35-5.
[Eighth International Symposium on Imprecise Probability: Theories adn Applications. Compiegne (FR), 02.07.2013-05.07.2013]
R&D Projects: GA ČR GA13-20012S
Grant - others:GA ČR(CZ) GAP403/12/2175
Program: GA
Institutional support: RVO:67985556
Keywords : marginal problem * belief function * algorithm * multidimensional model * convergence
Subject RIV: BA - General Mathematics
http://library.utia.cas.cz/separaty/2013/MTR/kratochvil-on open problems connected with application of the iterative proportional fitting procedure to belief functions.pdf

In probability theory, Iterative Proportional Fitting Procedure can be used for construction of a joint probability measure from a system of its marginals. The present paper studies a possibility of application of an analogous procedure for belief functions, which was made possible by the fact that there exist operators of composition for belief functions. In fact, two different procedures based on two different composition operators are introduced. The procedure based on the composition derived from the Dempster's rule of combination is of very high computationally complexity and, from the theoretical point of view, practically nothing is known about its behavior. The other one, which uses the composition derived from the notion of factorization, is much more computationally efficient, and its convergence is guaranteed by a theorem proved in this paper.
Permanent Link: http://hdl.handle.net/11104/0223481