0447270 20240103210553.120150925235959.9 84963972159 000377943800441 10.1109/EUSIPCO.2015.7362773 5 s. C cav_un_epca*0447269978-0-9928626-4-02076-14652236-2240NiceEURASIP2015 Approximate Bayesian computation ABC filtration cav_un_auth*0242543 Adaptivní systémy Department of Adaptive Systems AS AS Department of Adaptive Systems Dedecius Kamil UTIA-B Ústav teorie informace a automatizace AV ČR, v. v. i. http://library.utia.cas.cz/separaty/2015/AS/dedecius-0447270.pdf cav_un_auth*0303543 GP14-06678P GA ČR CZ Approximate Bayesian computation (ABC) filtration of state-space models replaces popular particle filters in cases where the observation models (i.e. likelihoods) are either computationally too demanding or completely intractable, but it is still possible to simulate from them. These sequential Monte Carlo methods evaluate importance weights based on the distance between the true observation and the simulated pseudo-observations. The paper proposes a new adaptive method consisting of probability kernel-based evaluation of importance weights with online determination of kernel scale. It is shown that the resulting algorithm achieves performance close to particle filters in the case of well-specified models, and outperforms generic particle filters and state-of-art ABC filters under heavy-tailed noise and model misspecification. cav_un_auth*0319341 23rd European Signal Processing Conference (EUSIPCO) 31.08.2015-04.09.2015 Nice FR BB 2016 1 PO RVO:67985556 http://hdl.handle.net/11104/0249574 1 S 2015 84963972159 SCOPUS 000377943800441 WOS cav_un_epca*0447269 Proceedings of 23rd European Signal Processing Conference 978-0-9928626-4-0 2076-1465 2236 2240 Nice EURASIP 2015