DESIGN PRINCIPLES FOR INDUCTIVE INFERENCE PROCEDURES

被引：0

作者：

Sergienko, I. V. ^{[1
]}

Gupal, A. M. ^{[1
]}

机构：

[1] Natl Acad Sci Ukraine, VM Glushkov Inst Cybernet, Kiev, Ukraine

来源：

CYBERNETICS AND SYSTEMS ANALYSIS | 2006年 / 42卷 / 04期

关键词：

recognition; inductive inference; error of a procedure; complexity of class of problems; learning sample; Bayesian procedure;

D O I：

10.1007/s10559-006-0089-0

中图分类号：

TP3 [计算技术、计算机技术];

学科分类号：

0812 ;

摘要：

The behavior of inductive inference procedures depending on the content of a learning sample is analyzed. It is shown that if a learning sample contains no information on some class of objects or statistical information on a priori probabilities of classes, then any procedure performs unpredictably badly and its error is strictly positive.

引用

页码：505 / 515

页数：11

共 16 条

[1] [Anonymous], 1973, PROBABILITY INFORM
[2] Asmus V. F., 1947, LOGIC
[3] Baldi P., 2001, BIOINFORMATICS MACHI
[4] Belikov V. B., 2005, DOKL AN, P10
[5] Optimal Pattern Recognition Procedures. Substantiation of Inductive Inference Procedures
A. M. Gupal
I. V. Sergienko
[J]. Cybernetics and Systems Analysis, 2003, 39 (1) : 27 - 32
[6] Gupal A. M., 2002, PROBL UPRAVL INFORM, P83
[7] Gupal A. M., 2001, PROBL UPRAVL INFORM, P5
[8] Efficiency of Bayesian classification procedure
Gupal, AM
Pashko, SV
Sergienko, IV
[J]. CYBERNETICS AND SYSTEMS ANALYSIS, 1995, 31 (04) : 543 - 554
[9] Lem N., 2004, SUMMA TECHNOLOGIAE
[10] Nemirovskii A.S., 1979, COMPLEXITY PROBLEMS

← 1 2 →