No free lunch for cross-validation

被引：37

作者：

Zhu, HY

Rohwer, R

机构：

[1] Neural Computing Research Group, Dept. of Comp. Sci. and Appl. Math., Aston University

来源：

NEURAL COMPUTATION | 1996年 / 8卷 / 07期

关键词：

D O I：

10.1162/neco.1996.8.7.1421

中图分类号：

TP18 [人工智能理论];

学科分类号：

081104 ; 0812 ; 0835 ; 1405 ;

摘要：

It is known theoretically that an algorithm cannot be good for an arbitrary prior. We show that in practical terms this also applies to the technique of ''cross-validation,'' which has been widely regarded as defying this general rule. Numerical examples are analyzed in detail. Their implications to researches on learning algorithms are discussed.

引用

页码：1421 / 1426

页数：6

共 5 条

[1] Theory of statistical estimation. [J].

Fisher, RA .

PROCEEDINGS OF THE CAMBRIDGE PHILOSOPHICAL SOCIETY, 1925, 22 :700-725

[2]

Ord-Hume A. W. J. G., 1977, PERPETUAL MOTION HIS

[3]

Wolpert D.H., 1995, SFITR9502010

[4] Bayesian regression filters and the issue of priors [J].

Zhu, HY ;

Rohwer, R .

NEURAL COMPUTING & APPLICATIONS, 1996, 4 (03) :130-142

[5] BAYESIAN INVARIANT MEASUREMENTS OF GENERALIZATION [J].

ZHU, HY ;

ROHWER, R .

NEURAL PROCESSING LETTERS, 1995, 2 (06) :28-31

← 1 →