讨论了具一个隐层单元的神经网络在鼠空间中逼近的特征性定理并给出了逼近估计.对于平移网络,建立了Favard型估计,Orlicz空间中的相应结果均作为应用而给出.
The characteristic theorem as well as the degree of approximating by neural networks and translation networks, with a single hidden layer in Bo spaces is established. The Favard type estimate is obtained as well. The correspondences of Orlicz spaces are given by one of the applications of Ba spaces.