Stationary and integrated autoregressive neural network processes

Stationary and integrated autoregressive neural network processes

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 overa...

Full description

Bibliographic Details
Main Authors: Trapletti, Adrian, Leisch, Friedrich, Hornik, Kurt
Format: Others
Language:en
Published: SFB Adaptive Information Systems and Modelling in Economics and Management Science, WU Vienna University of Economics and Business 1998
Subjects:
neuronales Netz / autoregressiver Prozess / Zeitreihenanalyse / Prognose
Online Access:http://epub.wu.ac.at/302/1/document.pdf
id ndltd-VIENNA-oai-epub.wu-wien.ac.at-epub-wu-01_228
record_format oai_dc
spelling ndltd-VIENNA-oai-epub.wu-wien.ac.at-epub-wu-01_2282017-02-28T05:22:38Z Stationary and integrated autoregressive neural network processes Trapletti, Adrian Leisch, Friedrich Hornik, Kurt neuronales Netz / autoregressiver Prozess / Zeitreihenanalyse / Prognose 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. SFB Adaptive Information Systems and Modelling in Economics and Management Science, WU Vienna University of Economics and Business 1998 Paper NonPeerReviewed en application/pdf http://epub.wu.ac.at/302/1/document.pdf Series: Working Papers SFB "Adaptive Information Systems and Modelling in Economics and Management Science" http://epub.wu.ac.at/302/
collection NDLTD
language en
format Others
sources NDLTD
topic neuronales Netz / autoregressiver Prozess / Zeitreihenanalyse / Prognose
spellingShingle neuronales Netz / autoregressiver Prozess / Zeitreihenanalyse / Prognose
Trapletti, Adrian
Leisch, Friedrich
Hornik, Kurt
Stationary and integrated autoregressive neural network processes
description 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. === Series: Working Papers SFB "Adaptive Information Systems and Modelling in Economics and Management Science"
author Trapletti, Adrian
Leisch, Friedrich
Hornik, Kurt
author_facet Trapletti, Adrian
Leisch, Friedrich
Hornik, Kurt
author_sort Trapletti, Adrian
title Stationary and integrated autoregressive neural network processes
title_short Stationary and integrated autoregressive neural network processes
title_full Stationary and integrated autoregressive neural network processes
title_fullStr Stationary and integrated autoregressive neural network processes
title_full_unstemmed Stationary and integrated autoregressive neural network processes
title_sort stationary and integrated autoregressive neural network processes
publisher SFB Adaptive Information Systems and Modelling in Economics and Management Science, WU Vienna University of Economics and Business
publishDate 1998
url http://epub.wu.ac.at/302/1/document.pdf
work_keys_str_mv AT traplettiadrian stationaryandintegratedautoregressiveneuralnetworkprocesses
AT leischfriedrich stationaryandintegratedautoregressiveneuralnetworkprocesses
AT hornikkurt stationaryandintegratedautoregressiveneuralnetworkprocesses
_version_ 1718417162586030080

Similar Items