RBF网络的学习过程与BP网络的学习过程类似，两者的主要区别在于各使用不同的作用函数。BP网络中隐层使用的是Sigmoid函数，其值在输入空间中无限大的范围内为非零值，因而是一种全局逼近的神经网络；而RBF网络中的作用函数是高斯基函数，其值在输入空间中有限范围内为非零值，因为RBF网络是局部逼近的神经网络。(The learning process of RBF networks is similar to the learning process of BP networks. The main difference between them is that they use different function functions. In the BP network layer using the Sigmoid function, the range of values in the input space within the infinite is nonzero, which is a global approximation of neural network; and the function of RBF network is the Gauss function, its value in the limited range of input space is nonzero, because the network is RBF the neural network of local approximation.
RBF network is a 3 layer feedforward network, from input to output mapping is nonlinear, and the hidden layer space to the output space mapping is linear, and the)