Optimal segmentation of experimental curves

A mathematical model is suggested of structural noise-like experimental curves in the form of a stochastic difference autoregression equation with a stepwise parameter vector. At each time parameter vector takes on one value from a finite set the succession of which is controlled by a Markov chain. The optimal curve segmentation problem is stated as that of designing a decision rule which minimizes the mean risk of noncoincidence of the actual and restored positions of switching times. Algorithms of optimal segmentation are developed for two different loss functions.