A multi-model adaptive predictor for stochastic processes with Markov switching parameters

From MaRDI portal





The paper deals with the derivation of an adaptive predictor (the multi- model adaptive predictor), using parallel processing techniques. The parameter vector, denoting unknown time-varying parameters of the process, is assumed to switch within a finite set of values. The switchings are modelled as a finite-state Markov chain whose transition probability matrix is assumed to be known. The whole adaptive system consists of a bank of minimum-square-error predictors, where the i-th predictor is designed on the assumption that the parameter vector is the i-th one of admissible values. The final prediction, which is optimal in the sense of Bayes, is obtained via a weighted sum over the individual predictors whose weighting factors are the a priori probabilities.











This page was built for publication: A multi-model adaptive predictor for stochastic processes with Markov switching parameters

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q3710434)