Sensitivity Analysis of the Split-Complex Valued Multilayer Perceptron Due to the Errors of the i.i.d. Inputs and Weights
- 4 September 2007
- journal article
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Neural Networks
- Vol. 18 (5), 1280-1293
- https://doi.org/10.1109/tnn.2007.894038
Abstract
In this paper, we analyze the sensitivity of a split-complex multilayer perceptron (split-CMLP) due to the errors of the inputs and the connection weights between neurons. For simplicity, all the inputs and weights studied here are independent and identically distributed (i.i.d.). To develop an algorithm to estimate the sensitivity of the entire split-CMLP, we compute statistically the sensitivity by using the central limit theorem (CLT). The results show that the sensitivity is affected by the number of the layers and the number of the neurons adopted in each layer. We derive a theoretical estimation of the sensitivity. Several numerical results of the sensitivity for the split-CMLP are presented, and they match the theoretical ones. The agreement between the theoretical results and experimental results verifies the feasibility of the proposed algorithm. Thus, we not only analyze the sensitivity of the split-CMLP due to the errors of the i.i.d. inputs and weights, but also develop an efficient algorithm to estimate the sensitivity.Keywords
This publication has 16 references indexed in Scilit:
- Computation of Adalines' Sensitivity to Weight PerturbationIEEE Transactions on Neural Networks, 2006
- Variance analysis of sensitivity information for pruning multilayer feedforward neural networksPublished by Institute of Electrical and Electronics Engineers (IEEE) ,2003
- Sensitivity analysis for input vector in multilayer feedforward neural networksPublished by Institute of Electrical and Electronics Engineers (IEEE) ,2002
- A sensitivity analysis algorithm for pruning feedforward neural networksPublished by Institute of Electrical and Electronics Engineers (IEEE) ,2002
- Sensitivity analysis of multilayer perceptron to input and weight perturbationsIEEE Transactions on Neural Networks, 2001
- Perturbation method for deleting redundant inputs of perceptron networksNeurocomputing, 1997
- Sensitivity analysis of single hidden-layer neural networks with threshold functions.IEEE Transactions on Neural Networks, 1995
- On the complex backpropagation algorithmIEEE Transactions on Signal Processing, 1992
- Sensitivity analysis of multilayer perceptron with differentiable activation functionsIEEE Transactions on Neural Networks, 1992
- The complex backpropagation algorithmIEEE Transactions on Signal Processing, 1991