Robust estimation of confidence interval in neural networks applied to time series

Rodrigo Salas, Romina Torres, Héctor Allende, Claudio Moraga

Research output: Chapter in Book/Report/Conference proceedingChapterpeer-review

2 Citations (Scopus)

Abstract

Artificial neural networks (ANN) have been widely used in regression or predictions problems and it is usually desirable that some form of confidence bound is placed on the predicted value. A number of methods have been proposed for estimating the uncertainty associated with a value predicted by a feedforward neural network (FANN), but these methods are computationally intensive or only valid under certain assumptions, which are rarely satisfied in practice. We present the theoretical results about the construction of confidence intervals in the prediction of nonlinear time series modeled by FANN, this method is based on M-estimators that are a robust learning algorithm for parameter estimation when the data set is contaminated. The confidence interval that we propose is constructed from the study of the Influence Function of the estimator. We demonstrate our technique on computer generated Time Series data.

Original languageEnglish
Title of host publicationLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
EditorsJose Mira, Jose R. Alvarez
PublisherSpringer Verlag
Pages441-448
Number of pages8
ISBN (Print)354040211X, 9783540402114
DOIs
Publication statusPublished - 2003

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume2687
ISSN (Print)0302-9743
ISSN (Electronic)1611-3349

Keywords

  • Confidence Interval
  • Feedforward Artificial Neural Networks
  • Robust Learning algorithm
  • Time Series

ASJC Scopus subject areas

  • Theoretical Computer Science
  • Computer Science(all)

Fingerprint

Dive into the research topics of 'Robust estimation of confidence interval in neural networks applied to time series'. Together they form a unique fingerprint.

Cite this