Stationary and integrated autoregressive neural network processes

Adrian Trapletti, Friedrich Leisch, Kurt Hornik

Publication: Working/Discussion PaperWU Working Paper

74 Downloads (Pure)

Abstract

We consider autoregressive neural network (ARNN) processes driven by additive noise. Sufficient conditions on the network weights (parameters) are derived for the ergodicity and stationarity of the process. It is shown that essentially the linear part of the ARNN process determines whether the overall process is stationary. A generalization to the case of integrated ARNN processes is given. Least squares training (estimation) of the stationary models and testing for non-stationarity are discussed. The estimators are shown to be consistent and expressions on the limiting distributions are given.

Publication series

SeriesWorking Papers SFB "Adaptive Information Systems and Modelling in Economics and Management Science"
Number24

WU Working Paper Series

  • Working Papers SFB \Adaptive Information Systems and Modelling in Economics and Management Science\

Cite this