非线性函数的自适应分区多神经网络学习及仿真实验

Adaptive Partitioning Learning of Nonlinear Function Using Neural Network and Simulation Experiment

根据BP神经网络学习非线性函数的精度与所学函数的区间大小及变化率等有关, 提出了一种非线性函数的自适应分区多神经网络学习方法, 这种方法根据学习精度的要求,自适应地把所学函数分成若干区间, 分别用一个BP神经网络去学习,从而使学习精度大大提高。 最后, 给出了学习一维函数和多维函数的仿真实例, 其结果表明分区学习的精度可提高10倍以上 。

This paper describes a method that a nonlinear function is learned with adaptive domain partitioning using neural networks , according as the accuracy of learning a nonlinear function using BP neural network has relation to dimension of the interval of learned function and its rate of change. Using this method, the nonlinear function to learn can be partitioned into many intervals with adaptive method according to requisition of learning accuracy using BP neural network , and the nonlinear function is learned in the each interval respectively using a BP neural network , thus, the accuracy of learning is greatly raised.