论文部分内容阅读
针对一类变参数 Sigmoid可调激活函数构成三层前向神经网络 ,分析其可调激活函数中参数所表示意义 ;给出了递进提升输出向量空间维数的可调变参数激活函数中参数选取的方法 ,解决了隐含神经元采用相同激活函数限制了神经网络逼近能力这一问题 .其目的给人们在采用变参数可调激活函数神经网络解决问题时 ,如何选取激活函数中的参数提供了一种数学依据和方法 .
For a class of variable parameter Sigmoid adjustable activation function, a three-layer feedforward neural network is constructed to analyze the meaning of the parameters in the adjustable activation function. The parameters in the activation function for the variable dimension parameter The method solves the problem that implicit neurons use the same activation function to limit the ability of neural network to approach.The purpose of this paper is to provide people with the parameters of activation function when using the neural network with variable parameter adjustable activation function to solve the problem A mathematical basis and method.