Dynamic Bayesian learning by expectation propagation

Tao Wei

doi:10.1117/12.664342

20 February 2006 Dynamic Bayesian learning by expectation propagation

Tao Wei

Proceedings Volume 6041, ICMIT 2005: Information Systems and Signal Processing; 60411N (2006) https://doi.org/10.1117/12.664342
Event: ICMIT 2005: Merchatronics, MEMS, and Smart Materials, 2005, Chongqing, China

Abstract

For modeling time-series data, it is natural to use directed graphical models, since they can capture the time flow. If arcs of a graphical model are all directed both within and between time-slice, the model is called dynamic Bayesian network (DBN). Dynamic Bayesian networks are becoming increasingly important for research and applications in the area of machine learning, artificial intelligence and signal processing. It has several advantages over other data analysis methods including rule bases, neural network, decision trees, etc. In this paper, there explored dynamic Bayesian learning over DBNs by a new deterministic approximate inference method called Expectation Propagation (EP). EP is an extension of belief propagation and is developed in machine learning. A crucial step of EP is the likelihoods recycling, which makes possible further improvement over the extended Kalman smoother. This study examined EP solutions to a non-linear state-space model and compared its performance with other inference methods such as particle filter, extended Kalman filter, etc.

Citation Download Citation

Tao Wei "Dynamic Bayesian learning by expectation propagation", Proc. SPIE 6041, ICMIT 2005: Information Systems and Signal Processing, 60411N (20 February 2006); https://doi.org/10.1117/12.664342

ACCESS THE FULL ARTICLE

INSTITUTIONAL
Select your institution to access the SPIE Digital Library.

SELECT YOUR INSTITUTION

PERSONAL
Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.

PERSONAL SIGN IN

No SPIE Account? Create one

PURCHASE THIS CONTENT

SUBSCRIBE TO DIGITAL LIBRARY

50 downloads per 1-year subscription

Members: $195

Non-members: $335 ADD TO CART

25 downloads per 1 - year subscription

Members: $145

Non-members: $250 ADD TO CART

PURCHASE SINGLE ARTICLE

Includes PDF, HTML & Video, when available