1: \begin{abstract}

2: We propose a general formalism of iterated random functions

3: with semigroup property, under which exact and approximate Bayesian 

4: posterior updates can be viewed as specific instances. 

5: A convergence theory for iterated random functions 

6: is presented. As an application of the general theory we analyze

7: convergence behaviors of exact and approximate message-passing

8: algorithms that arise in a sequential change point detection

9: problem formulated via a latent variable directed graphical model. 

10: The sequential inference algorithm and its supporting theory are 

11: illustrated by simulated examples. 

12: \end{abstract}