• Medientyp: E-Artikel
  • Titel: A Comparison of Some Error Estimates for Neural Network Models
  • Beteiligte: Tibshirani, Robert
  • Erschienen: MIT Press - Journals, 1996
  • Erschienen in: Neural Computation
  • Sprache: Englisch
  • DOI: 10.1162/neco.1996.8.1.152
  • ISSN: 0899-7667; 1530-888X
  • Entstehung:
  • Anmerkungen:
  • Beschreibung: <jats:p> We discuss a number of methods for estimating the standard error of predicted values from a multilayer perceptron. These methods include the delta method based on the Hessian, bootstrap estimators, and the “sandwich” estimator. The methods are described and compared in a number of examples. We find that the bootstrap methods perform best, partly because they capture variability due to the choice of starting weights. </jats:p>