BP神经网络模型中活化函数对网络性能影响的研究

RESEARCH ON THE INFLUENCE OF ACTIVATION FUNCTION ON NETWORKS ABILITIES IN BP NEURAL NETWORKS MODEL

本文在ＢＰ神经网络模型中，采用一种新型的神经元活化函数替换传统的Ｓ型活化函数，并在活化函数中引入了可调参量，研究表明，作这一替换后对提高网络的学习速度、抑制假饱和现象起到了很好的作用．

In this paper, S type neuron activation function in traditional BP neural networks model is replaced by a new type neuron activation function. Research results shows that this replacement helps accelerate network convergence speed and restrain false saturation.