Abstract
We propose a simple method that enhances the performance of Bayesian Regularization of Artificial Neural Network (ANN) through pre-training of initial network with the Early-Stopping algorithm. The proposed method is applied to the regularization of Feed-forward Neural Networks to regress three benchmark data series. Significant reduction in both the cross-validation error and the number of training over standard Bayesian Regularisation is achieved.
Similar content being viewed by others
References
MacKay, D. J. C.: A practical Bayesian framework for backprop networks, In: J. E. Moody, S. J. Hanson and R. P. Lippmann (eds), Advances in Neural Information Processing Systems, pp.839–846, 1992.
Vehtari, A. and Lampinen, J.: Bayesian neural networks for industrial applications. In: IEEE Midnight-Sun Workshop on Soft Computing Methods in Industrial Applications, Kuusamo, Finland, 1999.
Edwards, P., Peacock, A., Renshaw, D., Hannah, J.and Murray, A.: Minimizing risk using prediction uncertainty in neural network estimation fusion and its application to paper-making, IEEE Transactions on Neural Networks, 13(3) (2002), 726–731.
Gencay, R.and Min, Q.: Pricing and hedging derivative securities with neural networks: Bayesian regularization, early stopping, and bagging, IEEE Transactions on Neural Networks, 12(4) (2001), 726–734.
Bishop, C.: Neural Network for Pattern Recognition, Oxford University Press Inc., 1995.
Plate, T.: Controlling the hyperparameter search in MacKay 's Bayesian neural network Framework. In: G. B. Orr, K. R. Muller (eds), Neural Networks: Tricks of the Trade, Springer, pp. 93–112, 1998.
MacKay, D. J. C.: Bayesian methods for neural networks-FAQ, http://wol.ra.phy.cam. ac.uk/mackay/Bayes_FAQ.html, 2000.
Author information
Authors and Affiliations
Rights and permissions
About this article
Cite this article
Chan, Z.S.H., Ngan, H.W. & Rad, A.B. Improving Bayesian Regularization of ANN via Pre-training with Early-Stopping. Neural Processing Letters 18, 29–34 (2003). https://doi.org/10.1023/A:1026271406135
Issue Date:
DOI: https://doi.org/10.1023/A:1026271406135