On ``An improved approach for nonlinear system identification using neural networks''. (Q1420540)

[Concerns the paper by \textit{P. Gupta} and \textit{N. K. Sinha}, J. Franklin Inst. 336, 721--734 (1999; Zbl 0979.93508).] Neural networks are used for improving nonlinear system identification results. To this end, a learning algorithm with delta-bar-delta rule for multilayer feedforward neural networks is applied. It is shown that only the adaptation of one of the two parameters -- the learning rate and the activation function -- is sufficient. Experimental results that employ an adaptive \(\beta\) are presented in order to confirm the analysis.