‘Unobserved’ Monte Carlo Methods for Adaptive Algorithms
Many Signal Processing and Control problems are complicated by the presence of unobserved variables. Even in linear settings this can cause problems in constructing adaptive parameter estimators. In previous work the author investigated the possibility of developing an on-line version of so-called Markov Chain Monte Carlo methods for solving these kinds of problems. In this article we present a new and simpler approach to the same group of problems based on direct simulation of unobserved variables.
KeywordsMarkov Chain Monte Carlo Adaptive Algorithm
Unable to display preview. Download preview PDF.
- Kuk, A.Y.C., and Y. W. Cheng. (1997). The Monte Carlo Newton Raphson algorithm. Jl Stat Computation Simul.Google Scholar
- Kitagawa, G. (1998). Self organising state space model. Jl. Amer. Stat. Assoc, 93:1203–1215.Google Scholar
- Ng, L. and V. Solo. (2001). Errors-in-variables modelling in optical flow estimation. IEEE Trans. Im.Proc., to appear.Google Scholar
- Rappaport, T.S. (1996). Wireless Communication. Prentice Hall, New York.Google Scholar
- Solo, V. and X. Kong. (1995). Adaptive Signal Processing Algorithms. Prentice Hall, New Jersey.Google Scholar
- Solo, V. (1999). Adaptive algorithms and Markov chain Monte Carlo methods. In Proc. IEEE Conf Decision Control 1999, Phoenix, Arizona, IEEE.Google Scholar
- Solo, V. (2000a). ‘Unobserved’ Monte Carlo method for system identification of partially observed nonlinear state space systems, Part I: Analog observations. In Proc JSM2001, Atlanta, Georgia, August, page to appear. Am Stat Assocn.Google Scholar
- Solo, V. (2000b). ‘Unobserved’ Monte Carlo method for system identification of partially observed nonlinear state space systems, Part II: Counting process observations. In Proc 39th IEEE CDC, Sydney Australia. IEEE.Google Scholar
- Tekalp, M. (1995). Digital Video Processing. Prentice-Hall, Englewood Cliffs, N.J.Google Scholar
- Yakowitz, S. and F. Szidarovszky. (1984). A comparison of kriging with nonparametric regression methods. Jl Mult Anal.Google Scholar