Universal approximation of nonlinear system predictions in sigmoid activation functions using artificial neural networks

摘要

The sigmoid activation function cast-off to convert the equal of activation of units (neurons) in the output indicator. There are a numeral of mutual tasks in activation with the use of artificial neural networks (ANN). The maximum communal use of manifold functions to Multi Layered Perceptron (MLP) and the transmission of professions in research and engineering. However, given the wide range of problematic fields are applied in the MLP, it is interesting to suspect that the detailed difficulties that require one or exact activation utilities of the group. The aim of this paper is to consider the presentation of buildings MLP generalized who appeared deployment algorithm by numerous dissimilar functions to activate the sigmoid neurons of the hidden and output layers.