Optimal detection of changing properties of random sequences. I: Nonsequential detection (Q1103575)

An optimal Bayesian nonsequential algorithm of detection of a failure of a stochastic process observed at discrete instants of time is constructed. Recursion formulas are presented for the averaged likelihood ratio and the a posteriori probability of a failure, both of them being sufficient statistics. In case of independent observations these statistics are proved to be nonhomogeneous Markov functions. Examples are also presented. [For part II see the following review.]