A new algorithm of training neural networks by orthogonal weight functions and sensitivity analysis
Daiyuan Zhang · 2012
A new algorithm of training neural networks by orthogonal weight functions (OWFs) is proposed, which is based on the training algorithm using cubic spline weight functions. The weights obtained after training are orthogonal functions defined on the sets of input variables (input patterns). Sensitivity analyses for neural networks using OWFs are also discussed in this paper. The sensitivity formulae of OWFs neural networks are derived. Based on the analyses of sensitivity, theoretical sensitivity and approximation sensitivity are also proposed. Finally, the correctness of the results proposed in this paper is verified by computational simulations.