Number of the records: 1

Bayesian averaging of regressive models

SYSNO ASEP	0346948
Document Type	C - Proceedings Paper (int. conf.)
R&D Document Type	Conference Paper
Title	Bayesian averaging of regressive models
Author(s)	Dedecius, Kamil (UTIA-B)_{RID, ORCID} Jirsa, Ladislav (UTIA-B)_RID Pištěk, Miroslav (UTIA-B)_{RID, ORCID}
Source Title	Proceedings of the 11th International PhD Workshop on Systems and Control. - Veszprém, Maďarsko : Faculty of Information Technology, University of Pannonia, 2010 - ISBN 978-615-5044-00-7
Pages	s. 1-6
Number of pages	6 s.
Action	11th International PhD Workshop on Systems and Control a Young Generation Viewpoint
Event date	01.09.2010-03.09.2010
VEvent location	Veszprém
Country	HU - Hungary
Event type	EUR
Language	eng - English
Country	HU - Hungary
Keywords	bayesian modelling ; model averaging ; estimation
Subject RIV	JD - Computer Applications, Robotics
R&D Projects	7D09008 GA MŠMT - Ministry of Education, Youth and Sports (MEYS)
CEZ	AV0Z10750506 - UTIA-B (2005-2011)
Annotation	In the real world, it is often possible to model certain variables using several different regressive models. However, as the theoretical (mathematical, physical...) description of the real world is almost never perfect, there exists uncertainty about the true or best-fitting model. This paper deals with an issue of `mixing' information from multiple potentially true models, which run in parallel, to obtain a single outcome, taking the model uncertainty into account. While this issue has been addressed by many research papers in the past, most of them were developed for the static cases or for the state-space models. Here, we discuss an enhancement for a class of input-output regressive models. The described method allows to switch among models to reflect their modelling performance. If there is a single best model, the method quickly converges to it.
Workplace	Institute of Information Theory and Automation
Contact	Markéta Votavová, votavova@utia.cas.cz, Tel.: 266 052 201.
Year of Publishing	2011

Number of the records: 1