BriefGPT.xyz
Jan, 2020
光滑函数的深度网络逼近
Deep Network Approximation for Smooth Functions
HTML
PDF
Jianfeng Lu, Zuowei Shen, Haizhao Yang, Shijun Zhang
TL;DR
本文研究了深度修正线性单元网络关于宽度和深度同时逼近平滑函数的最优逼近误差特性,并且证明了多元多项式可以被宽度为O(N)和深度为O(L)的深ReLUNetwork逼近,而且证明了具有O(N lnN)宽度和O(L lnL)深度的深ReLUNetwork能够用近乎最优的逼近误差逼近 f∈ C^s ([0,1]^d)。
Abstract
This paper establishes optimal
approximation error
characterization of
deep relu networks
for
smooth functions
in terms of both
→