距离空间中插值神经网络的误差估计

ESTIMATIONS OF ERROR FOR INTERPOLATION NEURAL NETWORKS IN DISTANCE SPACES

研究距离空间中的神经网络插值与逼近问题.首先引进一类广义的激活函数,用比较简洁的方法讨论距离空间中插值神经网络的存在性,然后给出插值神经网络逼近连续函数的误差估计.

This paper deals with the interpolation and approximation of neural networks in distance spaces. First a class of generalized activation function and is introduced a simple method is used to prove the existence of interpolation networks in the distance spaces. And then the estimates of error for the interpolation neural networks approximating continuous function is given.