Recursive Estimation of Mixtures of Exponential and Normal Distributions

0448117 - ÚTIA 2016 RIV PL eng C - Konferenční příspěvek (zahraniční konf.)
Suzdaleva, Evgenia - Nagy, Ivan - Mlynářová, Tereza
Recursive Estimation of Mixtures of Exponential and Normal Distributions.
Proceedings of the 2015 IEEE 8th International Conference on Intelligent Data Acquisition and Advanced Computing Systems: Technology and Applications (IDAACS). Piscataway: IEEE, 2015, s. 137-142. ISBN 978-1-4673-8361-5.
[International Conference on Intelligent Data Acquisition and Advanced Computing Systems: Technology and Applications /8./ (IDAACS'2015). Warsaw (PL), 24.09.2015-26.09.2015]
Grant CEP: GA MŠMT(BE) 7H14005; GA MŠMT 7H14004; GA ČR(CZ) GA15-03564S
Institucionální podpora: RVO:67985556
Klíčová slova: recursive mixture estimation * mixture of different distributions * dynamic switching model * exponential distribution
Kód oboru RIV: BB - Aplikovaná statistika, operační výzkum
http://library.utia.cas.cz/separaty/2015/ZS/suzdaleva-0448117.pdf

The paper deals with estimation of a mixture of normal and exponential distributions with the dynamic model of their switching. A separate estimation of normal or exponential mixtures is solved by various approaches in many papers over the world. However, in some application areas, data are of such a nature that they should be described by a combination of exponential and normal models. The paper proposes a recursive Bayesian algorithm of estimation of such a mixture based on continuously measured data. Specific tasks the paper solves are: (i) parameter estimation of both the types of components; (ii) parameter estimation of the dynamic switching model and (iii) detection of the currently active component. Results of experiments with real data are demonstrated.
Trvalý link: http://hdl.handle.net/11104/0250068