用BP神经网络实现函数逼近的方法及其讨论

The Realization and Discussion of a Function's Approximation Based on BP Neural Networks

构造一个三层BP神经网络,实现了连续函数的逼近。讨论了函数值在非[0,1]区间时,样本的归一化问题,提出了新的变换关系式,加快了算法的收敛速度。

This paper introduced a three layer BP neural network,and realized the approximation of a continuous function.Based on the above BP neural network,A new transform normalized the function's output are put forward to and the training speed is fastened,which is proved by the result.