机器学习中知识动态获取在函数逼近中的探究

The dynamic acquisition of knowledge in machine learning inquiry in the function approximation

本文将机器学习中的知识动态获取体现在神经网络的方法中,从而研究神经网络的函数逼近方法,首先分析了神经网络在函数逼近中应用的相关理论,然后将BP神经网络应用于函数的逼近中并通过实验得到理想的效果。最后本文首次提出将GRNN(广义回归神经网络)运用于实际的函数逼近之中,得到了误差极小(接近于零)的完美逼近结果,并且通过实验验证了该神经网络训练速度快和非线性映射能力很强的优点。

This paper puts dynamic acquisition of knowledge in machine learning into neural network approach, then study neural networks about the methods of function approximation, firstly analyzes some relevant theories about the neural networks in function approximation. And then putting BP neural networks into function approximation and getting the desired results through experiments.Finally, applied the GRNN （Generalized Regression neural networks） to the actual function approximation, getting the result of function approximation in which error has been minimal. And through experiment verify this neural network which has the advantage of raining high speed and strong non-linear mapping capability.