Weighted Probabilistic Opinion Pooling Based on Cross-Entropy

0450905 - ÚTIA 2016 RIV CH eng C - Conference Paper (international conference)
Sečkárová, Vladimíra
Weighted Probabilistic Opinion Pooling Based on Cross-Entropy.
Neural Information Processing. Cham: Springer International Publishing, 2015 - (Sabri, A.; Tingwen, H.; Weng, K.; Qingshan, L.), s. 623-629. Lecture Notes in Computer Science, 9490. ISBN 978-3-319-26534-6. ISSN 0302-9743.
[22nd International Conference on Neural Information Processing (ICONIP2015). Istanbul (TR), 09.11.2015-12.11.2015]
R&D Projects: GA ČR GA13-13502S
Institutional support: RVO:67985556
Keywords : Minimum cross-entropy principle * Kullback-Leibler divergence * Linear opinion pooling * Combining probability distributions
Subject RIV: BC - Control Systems Theory
http://library.utia.cas.cz/separaty/2015/AS/seckarova-0450905.pdf

In this work we focus on opinion pooling in the finite group of sources introduced in [Seckarova, 2015]. This approach, heavily exploiting Kullback-Leibler divergence (also known as cross-entropy), allows us to combine sources’ opinions given in probabilistic form, i.e. represented by the probability mass function (pmf). However, this approach assumes that sources are equally reliable with no preferences on, e.g., importance of a particular source. The discussion about the influence of the combination by preferences among sources (represented by weights) and numerical demonstration of the derived theory on an illustrative example form the core of this contribution.
Permanent Link: http://hdl.handle.net/11104/0252663