Analysis and expansion of a neural architecture capable of computing its own reliability

• Abner C. Rodrigues Neto UFSC
• Mauro Roisenberg UFSC
• Guenther Schwedersky Neto Petrobras

Abstract

There are several ways to calculate a measure of confidence to the output of neural networks, but in general these approaches require some restrictions that are not always observed in real problems or even not provide a measure of performance that guarantees the desired level of confidence or which do not reflect the distribution of training data. This paper analyzes and extends a model of neural network that calculates the confidence of its outputs, the Validity Index Network, we remove it restrictions in the calculation of the density and improve the probability coverage of the prediction levels when the training data have variable density.

References

Bishop, C. M. (1994). Novelty detection and neural network validation. In IEE Proc.-Vis. Image Signal Process., Vol. 141.

Bishop, C. M. (1995). Neural Netowrks for Pattern Recognition. Oxford Press.

Chinman, R. and Ding, J. (1998). Prediction limit estimation for neural network models. Neural Networks, IEEE Transactions on, 9(6):1515–1522.

Chryssolouris, G., Lee, M., and Ramsey, A. (1996). Confidence interval prediction for neural network models. In IEEE Trans. Neural Networks 7, pages 229–232.

Hassoun, M. H. (1995). Fundamentals of Artificial Neural Networks. MIT Press.

Holmstrom, L. and Hamalainen, A. (1993). The self-organizing reduced kernel density estimator. In IEEE International Conference on Neural Networks, pages 417–421.

Hwang, J. T. G. and Ding, A. A. (1997). Prediction intervals for artificial neural networks. In J. American Statistical Association 92(438), pages 748–757.

Leonard, J. A., Kramer, M. A., and Ungar, L. H. (1992). A neural network architeture that compute its own reliability. In Computers Chem. Engng.

Moody, J. and Darken, C. J. (1989). Fast learning in networks of locally tuned processing units. In Neural Computation 1, pages 281–294.

Parzen, E. (1962). On estimation of a probability density function and mode ann. In Math. Statist. 33, pages 1065–1076.

Shao, R., Martin, E. B., Zhang, J., and Morris, A. J. (1997). Confidence bounds for neural network representations. In Computers chem. Engng 21, pages 1173–1178. Elsevier Science Ltd.

Terrell, G. R. and Scott, D. W. (1992). Variable kernel density estimation. In The Annals of Statistics, Vol. 20, No. 3., pages 1236–1265. Institute of Mathematical Statistics.

Veaux, R. D., Schweinsberg, J., and Shellington, D. (1998). Prediction intervals for neural networks via nonlinear regression. In Computers chem. Engng 21, pages 1173–1178. Elsevier Science Ltd.

Wedding, D. K. and Cios, K. J. (1997). Cetainty factors versus pazen windows as reliability measures in rbf networks. In Neurocomputing 19, pages 151–165. Elsevier Ltd.

Yang, L., Kavli, T., Carlin, M., Clausen, S., and de Groot, P. (1991). An evaluation of confidence bound estimation methods for neural networks. In ESIT 2000.