Analysis of the Relationship between Partially Dynamic Bayesian Network Architecture and Inference Algorithm Effectiveness
|
Permanent citation URL:
|
Files in This Item:
| Title: | Analysis of the Relationship between Partially Dynamic Bayesian Network Architecture and Inference Algorithm Effectiveness |
| Author(s): | Cannon, Stephen J. |
| Keywords: | Bayesian; Network; PDBN; BN Generator; SPI; Boyen-Koller |
| Issue Date: | 24-Jul-2008 |
| Abstract: | This thesis examines the relationship between the architecture of partially dynamic Bayesian networks and the effectiveness of various inference algorithms using these Bayesian networks. The algorithms studied were the symbolic probabilistic inference algorithm, the particle filter inference algorithm, and Boyen-Koller inference algorithm. The purpose of this research is to provide empirical support for theoretical models of the speed and accuracy of each of these inference algorithms as well as to develop statistical models that utilize computationally and conceptually simple factors. The author shows that the empirical results for the speed of inference of each inference algorithm generally agrees with the theoretical complexity models of each algorithm. The author also developed empirical models that predict the variance of speed of each of the inference algorithms explored in this research. |
| URI: | http://hdl.handle.net/1920/3181 |
| Area Of Study: | Systems Engineering |
| Institution: | George Mason University |
| Degree Level: | master's |
| Degree Name: | Master of Science in Systems Engineering |
| Appears in Collections: | The Volgenau School of Information Technology and Engineering |
Items in MARS are protected by copyright, with all rights reserved, to the extent allowed by law.