Empirical bayes linear regression with unknown model order
2007 (English)In: 2007 IEEE International Conference on Acoustics, Speech, and Signal Processing, Vol III, Pts 1-3, Proceedings, IEEE , 2007, 773-776 p.Conference paper (Refereed)
We study the maximum a posteriori probability model order selection algorithm for linear regression models, assuming Gaussian distributed noise and coefficient vectors. For the same data model, we also derive the minimum mean-square error coefficient vector estimate. The approaches are denoted BOSS (Bayesian Order Selection Strategy) and BPM (Bayesian Parameter estimation Method), respectively. Both BOSS and BPM require a priori knowledge on the distribution of the coefficients. However, under the assumption that the coefficient variance profile is smooth, we derive "empirical Bayesian" versions of our algorithms, which require little or no information from the user. We show in numerical examples that the estimators can outperform several classical methods, including the well-known AIC and BIC for order selection.
Place, publisher, year, edition, pages
IEEE , 2007. 773-776 p.
, ICASSP, IEEE International Conference on Acoustics, Speech and Signal Processing - Proceedings, ISSN 1520-6149 ; 3
Bayes procedures, Least mean square methods, Linear systems, Modeling, Parameter estimation
Telecommunications Electrical Engineering, Electronic Engineering, Information Engineering
IdentifiersURN: urn:nbn:se:kth:diva-155598DOI: 10.1109/ICASSP.2007.366794ISI: 000248906600194ScopusID: 2-s2.0-34547494566ISBN: 1424407281ISBN: 978-142440728-6OAI: oai:DiVA.org:kth-155598DiVA: diva2:765088
2007 IEEE International Conference on Acoustics, Speech and Signal Processing, ICASSP '07, 15 April 2007 through 20 April 2007, Honolulu, HI, United States
QC 201411212014-11-212014-11-072014-11-21Bibliographically approved