Sensitivity Analysis of Madalines to Weight Perturbation
This paper aims at exploring the behavior of the sensitivity for an ensemble of Madalines. An algorithm is first given to compute the Madalines’ sensitivity, and its efficiency is verified by computer simulations. Then, based on the algorithm, the sensitivity analysis is conducted, which shows that the dimension of input has little effect on the sensitivity as long as the dimension is sufficient large, and the increases in the number of Adalines in a layer and the number of layers will lead the sensitivity to increase under an upper bound. The analysis results will be useful for designing robust Madalines.
KeywordsNeural Network Weight Vector Input Vector Feedforward Neural Network Output Deviation
Unable to display preview. Download preview PDF.
- 3.Fu, L., Chen, T.: Sensitivity Analysis for Input Vector in Multilayer Feedforward Neural Networks. In: Proc. of IEEE Int. Conf. on Neural Networks, vol. 1, pp. 215–218 (1993)Google Scholar
- 8.Wang, Y., Zeng, X., Han, L.: Sensitivity of Madalines to Input and Weight Perturbations. In: Proceedings of IEEE International Conference on Machine Learning and Cybernetics, November 2003, pp. 1349–1354 (2003)Google Scholar
- 9.Gong, J., Zhao, G.: An approximate algorithm for bivariate normal integral. Computational Structural Mechanics and Applications 13(4), 494–497 (1996)Google Scholar