详细信息
l^P-系数正则化Shannon采样学习算法收敛速度(英文) 被引量:1
The Learning Rate of l^p-coefficient Regularized Shannon Sampling Algorithm
文献类型:期刊文献
中文题名:l^P-系数正则化Shannon采样学习算法收敛速度(英文)
英文题名:The Learning Rate of l^p-coefficient Regularized Shannon Sampling Algorithm
作者:盛宝怀[1]
机构:[1]绍兴文理学院数学系
年份:2014
卷号:0
期号:6
起止页码:905
中文期刊名:数学进展
外文期刊名:Advances in Mathematics
收录:北大核心2011、北大核心、CSCD、CSCD2013_2014
基金:Supported by NSFC(No.10871226,No.61179041)
语种:中文
中文关键词:Shannon采样算法;学习理论;系数正则化;学习速度
外文关键词:Shannon sampling algorithm;;learning theory;;coefficient regularization;;learning rate
中文摘要:研究l^P-系数正则化意义下Shannon采样学习算法的收敛速度估计问题.借助l^P-空间的凸性不等式给出了样本误差和正则化误差的上界估计,并给出了用K-泛函表示的逼近误差估计.将K-泛函的收敛速度估计转化为平移网络逼近问题,在此基础上给出了用概率表示的学习速度.
外文摘要:In the present paper,we provide an investigation on the learning rate of the Shannon sampling algorithms with l^p-coefficient regularization.We give the upper bounds for the sample error and the regularization error with the convex inequality in l^p-spaces and show the approximation error by a K-functional whose convergence rate can be sum up to the translation network approximation.Basing on these estimates we show an explicit learning rate in possibility.
参考文献:
正在载入数据...