Journal of Automation and Information Sciences

ISSN for PRINT: 1064-2315

For Online Access

Best Paper Award Selection - Editorial Board Site

2006, Volume38

Issue 11

83 pages

DOI: 10.1615/J Automat Inf Scien.v38.i11

Issue price - $238.00

Boris A. Beletskiy
V.M. Glushkov Institute of Cybernetics of National Academy of Sciences of Ukraine, Kiev, Ukraine

Alexandra A. Vagis
V.M. Glushkov Institute of Cybernetics of National Academy of Sciences of Ukraine, Kiev, Ukraine

Sergey V. Vasilyev
V.M. Glushkov Institute of Cybernetics of National Academy of Sciences of Ukraine, Kiev, Ukraine

Nikita A. Gupal
National Technical University of Ukraine “Kiev Polytechnical Institute”, Kiev, Ukraine

Behavior of inductive procedures depending on content of learning sampling is studied. We demonstrate, that if the learning sampling contains no information about some class of objects or statistical information about a priori probabilities of classes, then any procedure works badly and its error is strictly positive. An estimate of error of Bayesian recognition procedure depending on size of learning sampling and other parameters is derived. Suboptimality of Bayesian approach is proved, complexity of class of problems is assessed.

DOI: 10.1615/J Automat Inf Scien.v38.i11.60 Download article, 56-73 pages

Article price - $50.00

<< Previous article   Next article >>