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。

RBF网络是一种3层前向网络，由输入到输出的映射是非线性的，而隐层空间到输出空间的映射是线性的，而且RBF网络局部逼近的神经网络，因而采用RBF网络大大加快学习速度并避免局部极小问题，适合于实时控制的要求。采用RBF网络构成神经网络控制方案，可有效提高系统的精度、鲁棒性和自适应性。

-RBF network learning process and the learning process is similar to the BP network, the main difference is that each use a different role functions. BP network hidden layer using a Sigmoid function, its value in the input space of infinite range of non-zero value, which is a global approximation of the neural network　and the role of RBF network function is a Gaussian basis functions, which value in the input space within a limited range of non-zero value, because RBF network is a local approximation of the neural network. RBF network is a three-layer forward network, the mapping　input to output is nonlinear, while the hidden layer space to the output space mapping is linear, RBF network and local approximation of the neural network, which greatly accelerated learning using RBF network speed and avoid local minima problem, suitable for real-time control. Using RBF network of neural network control scheme can effectively improve the accuracy of the system, robustness and adaptability.