BP神经网络在多维非线性函数拟合中的应用

为了解决常规方法拟合多维非线性函数的预测结果精度较低问题,采用非线性映射能力较强的BP神经网络进行多维非线性函数拟合。通过对BP神经网络结构及训练过程的分析,依据给定的两维非线性函数构建BP神经网络模型,并利用MATLAB软件对BP神经网络进行预测。仿真结果表明,BP神经网络对于两维非线性函数有良好的拟合能力,拟合误差小、收敛速度快,也使该方法具有较好的推广性。

The Application of BP Neural Network in Multidimensional Nonlinear Function

In order to solve the low precision prediction problem in conventional method of multidimensional nonlinear function fitting, the strong non-linear mapping of BP neural network in multidimensional nonlinear function fitting is applied. By analyzing the BP neural network structure and training process, BP neural network model is established by the given two dimensional nonlinear, and prediction about BP neural network is made with MATLAB software. The simulations reveal that the application of BP neural network in two dimensional nonlinear function enjoys higher accuracy of fitting and fast convergence speed, which make it possible to generalize it.