噪声影响的泛空间上的学习理论关键定理  

Key theorem of learning theory with samples corrupted by noise on pan-space

作　　者：李俊华[1] 高林庆[2] 李海军[3] 

机构地区：[1]河北大学数学与计算机学院,河北保定071002 [2]河北大学研究生学院,河北保定071002 [3]中国地质大学长城学院,河北保定071002

出　　处：《计算机工程与应用》2012年第27期49-52,共4页Computer Engineering and Applications

基　　金：河北省教育厅资助科研项目(No.2010109)

摘　　要：关键定理是统计学习理论的重要组成部分,但目前其研究主要集中在概率空间上且假设样本不受噪声的影响。鉴于此,提出了泛空间上样本受噪声影响的期望风险泛函、经验风险泛函以及经验风险最小化原则的定义,给出并证明了泛空间上样本受噪声影响的学习理论的关键定理。The key theorem plays an important role in the statistical learning theory. However, the researches about it at present focus on probability space and the samples are supposed to be noise-free. In consideration of these facts, the definitions of expected risk functional, empirical risk functional and empirical risk minimization principle with samples corrupted by noise on pan-space are proposed, and the key theorem of learning theory with samples corrupted by noise on pan-space is proposed and proved.

关 键 词：泛空间 泛随机变量 噪声 经验风险最小化原则 关键定理 

分 类 号：TP181[自动化与计算机技术—控制理论与控制工程]