两大类典型前馈神经网络的性能比较研究

The perfotmance comarison study of two major typical categories of feed-forward neural network

介绍了BP及RBF神经网络的基本理论,阐述了这两类算法的基本原理,其后从理论上对二者的性能进行了比较,最后在MATLAB环境下对具体的实例进行了仿真,结果表明RBF网络的泛化能力在多方面都优于BP网络,但是在解决具有相同精度要求的问题时,BP网络的结构要比RBF网络简单,因此,在应用中可以以此来指导神经网络的设计。

The basic theory of BP and RBF neural network having been introduced, and described the basic principles of the two algocithms, then the performance of the two algorithms are compared, in theory, at last researching on the simulating of specific examples in the MATLAB environment. Simulation results show that RBF network generalization ability is superior in many aspects to BP network, but when solve with the same accuracy problem, the structure of BP network is simple than the RBF network which can guide the design.