arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2012.08443cs.LGcs.NAmath.NAmath.STstat.TH
收藏

基于随机初始化的神经网络训练的整体误差强分析

Strong overall error analysis for the training of artificial neural networks via random initializations

University of Münster(明斯特大学)

Arnulf Jentzen, Adrian Riekert

AI总结:

本文在深度监督学习中改进了整体误差收敛估计,证明在任意独立同分布随机初始化的随机优化算法下,网络深度只需以更慢速度增长即可达到相同逼近速率。

摘要:

尽管基于深度学习的逼近算法已在众多问题中得到非常成功的应用,但目前从数学角度看,其性能背后的原因尚未被完全理解。近期,在深度监督学习情境下已获得整体误差收敛的估计,但收敛速度极慢。本文部分改进了这些估计。更具体地说,我们证明,为了获得相同的逼近速率,神经网络的深度只需以远慢于此前要求的速度增长。该结果适用于任意具有独立同分布随机初始化的随机优化算法。

英文摘要:

Although deep learning based approximation algorithms have been applied very successfully to numerous problems, at the moment the reasons for their performance are not entirely understood from a mathematical point of view. Recently, estimates for the convergence of the overall error have been obtained in the situation of deep supervised learning, but with an extremely slow rate of convergence. In this note we partially improve on these estimates. More specifically, we show that the depth of the neural network only needs to increase much slower in order to obtain the same rate of approximation. The results hold in the case of an arbitrary stochastic optimization algorithm with i.i.d.\ random initializations.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑