作者提出了一种新的 BP神经网络模型 ,其隐层激活函数采用中心参数可调的 Gaussian函数 ,输出层采用斜度可调的 Sigmoid函数 ,从而神经元具有了更强的信息存储、处理能力。由于采用组合函数 ,将 Gaussian函数良好的局部性和 Sigmoid函数良好的全局性相结合 ,提高了神经网络的收敛速度。几个典型实验的结果表明 ,与传统 BP网络模型相比 ,新网络模型在学习能力和泛化推广能力方面都有明显提高 The authors propose a new BP neural network model. The hidden layer activation function adopts a Gaussian function with adjustable center parameters, and the output layer adopts a Sigmoid function with adjustable slope. Therefore, neurons have stronger information storage and processing capabilities . Due to the combination function, the good locality of Gaussian function and the good globality of Sigmoid function are combined to improve the convergence speed of neural network. The results of several typical experiments show that compared with the traditional BP network model, the new network model has obviously improved both in learning ability and generalization ability