A Study of Online Function Approximation System Based on BP Neural Network

Neural network is one of the most popular computational model, it is widely used in system identification, machine learning and artificial intelligence areas. In this paper, an online function approximation system is designed by back propagation (BP) neural network. In addition, the dynamic error, static error and tracking sensitivity of the system are studied to evaluate the performance of the system. The results indicate that the system has the capability to approximate any high-order function.