MODEL DISCRIMINATION USING AN ALGORITHMIC INFORMATION CRITERION

被引：17

作者：

MACIEJOWSKI, JM ^{[1
]}

机构：

[1] UNIV CAMBRIDGE PEMBROKE COLL,CAMBRIDGE,ENGLAND

来源：

AUTOMATICA | 1979年 / 15卷 / 05期

关键词：

Computability; economic systems; identification; information theory; Kolmogorov complexity; modelling; philosophical aspects; prediction; recursive functions; social and behavioral sciences; water pollution;

D O I：

10.1016/0005-1098(79)90006-2

中图分类号：

TP [自动化技术、计算机技术];

学科分类号：

0812 ;

摘要：

A method of choosing between competing models of small sets of observations is presented. The intended application is to the modelling of 'badly defined' systems. The ideas of algorithmic information theory are surveyed, and applied to model discrimination. A novel definition of a model as a program which computes the observations exactly is given; this allows the size of a model to be interpreted as a measure of the informativeness of the model. It also imposes a trade-off between approximation and complexity, which is essential if overfitting of models is to be avoided. Two examples of discriminating between models of field data are presented. © 1979.

引用

页码：579 / 593

页数：15

共 31 条

[1]

Barzdin J. M., 1972, SOV MATH DOKL, V13, P1224

[2]

BECK MB, 1978, RR7819 IIASA RES REP

[3]

BECK MB, 1976, J ENVIRON ENG DIV-AS, V102, P909

[4] TOWARD A MATHEMATICAL THEORY OF INDUCTIVE INFERENCE [J].

BLUM, L ;

BLUM, M .

INFORMATION AND CONTROL, 1975, 28 (02) :125-155

[5]

BOX GEP, 1970, TIME SERIES ANAL FOR

[6] ON LENGTH OF PROGRAMS FOR COMPUTING FINITE BINARY SEQUENCES [J].

CHAITIN, GJ .

JOURNAL OF THE ACM, 1966, 13 (04) :547-+

[7]

CHALLIS MF, 1975, ALGOL W LANGUAGE SPE

[8]

Deutsch R, 1965, ESTIMATION THEORY

[9] SYSTEM IDENTIFICATION, APPROXIMATION AND COMPLEXITY [J].

GAINES, BR .

INTERNATIONAL JOURNAL OF GENERAL SYSTEMS, 1977, 3 (03) :145-174

[10]

Ivakhnenko A. G., 1977, Soviet Automatic Control, V10, P24

← 1 2 3 4 →