Convergence and asymptotic behaviour of parallel algorithms

Publisher Website
Google Scholar
Abstract
This paper analysis the quality of bayesian estimates when the system dynamics and noise statistics are unknown. Emphasis is put on the evaluation with time of estimates covariance in open and closed loop so that valuable information can be obtained on the convergence and accuracy of the bayesian identification algorithm. Analytical expressions for the asymptotic behaviour of the estimates are obtained. Hence a clear relationship appears between the precision of the estimates, the number of data, the system dynamics and noise environment. In closed-loop bayesian identification the relation between quality of the identification and quality of control appears (dual effect). The theoretical results are tested on discrete examples and numerical results of simulation are given to confirm them.

This publication has 13 references indexed in Scilit: