A Comparison of Some Error Estimates for Neural Network Models

@article{Tibshirani1996ACO,
  title={A Comparison of Some Error Estimates for Neural Network Models},
  author={R. Tibshirani},
  journal={Neural Computation},
  year={1996},
  volume={8},
  pages={152-163}
}
  • R. Tibshirani
  • Published 1996
  • Computer Science
  • Neural Computation
  • 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. 

    Topics from this paper.

    Confidence estimation methods for neural networks : a practical comparison
    • 156
    • Highly Influenced
    • Open Access
    Practical Confidence and Prediction Intervals
    • 273
    • Open Access
    Construction of confidence intervals for neural networks based on least squares estimation
    • 77
    • Open Access
    Estimations of error bounds for neural-network function approximators
    • 40
    Evaluating Neural Network Predictors by Bootstrapping
    • 55
    • Open Access
    Machine learning approaches for estimation of prediction interval for the model output
    • 205
    A Bootstrap Evaluation of the Effect of Data Splitting on Financial Time Series
    • 92
    • Open Access

    References

    Publications referenced by this paper.
    SHOWING 1-10 OF 33 REFERENCES
    Introduction to the theory of neural computation
    • 7,511
    • Open Access
    A Practical Bayesian Framework for Backpropagation Networks
    • 2,052
    • Open Access
    Evaluating Neural Network Predictors by Bootstrapping
    • 55
    • Open Access
    A Bootstrap Evaluation of the Effect of Data Splitting on Financial Time Series
    • 92
    • Open Access
    Statistical Models in S
    • 2,154
    Connectionist Learning Procedures
    • 1,415
    • Open Access
    Pattern classification using neural networks
    • 668