Bayesian network management |
| |
Authors: | Ganesh Ayalvadi Green Peter O'Connell Neil Pitts Susan |
| |
Institution: | (1) BRIMS, HP Labs, Filton Road, Stoke Gifford, Bristol, BS12 6QZ, UK E-mail:;(2) Department of Mathematics, University of Bristol, Bristol, BS8 1TW, UK;(3) Statistical Laboratory, University of Cambridge, Cambridge, CB2 1SB, UK E-mail: |
| |
Abstract: | We formulate some general network (and risk) management problems in a Bayesian context, and point out some of the essential
features. We argue and demonstrate that, when one is interested in rare events, the Bayesian and frequentist approaches can
lead to very different strategies: the former typically leads to strategies which are more conservative. We also present an
asymptotic formula for the predictive probability of ruin (for a random walk with positive drift) for large initial capital
and large number of past observations. This is a preliminary investigation which raises many interesting questions for future
research.
This revised version was published online in June 2006 with corrections to the Cover Date. |
| |
Keywords: | Bayesian decision theory large deviations effective bandwidths traffic characterization telecommunication networks risk theory gambler's ruin |
本文献已被 SpringerLink 等数据库收录! |
|