UDC 004.942+623.454.862

Linear model selection criteria for the solution of discrete ill-posed problems on the basis of singular value decomposition and random projection

/ E.G. Revunova // Kibernetika i sistemnyi analiz. — 2016. — Vol. 52, N 4. — P. 174–192.

The criteria are developed to determine the optimal number of linear model components for the solution of discrete ill-posed problem by truncated singular value decomposition and the method of random projection. To develop the model selection criteria, we studied the behavior of solution vector error and recovered right side vector error versus the dimension of the model and their minima. The results of the experimental investigation of the developed criteria are also provided. Figs: 3. Tabl.: 2. Refs.: 45 назв.

Keywords:

discrete ill-posed problem, truncated singular value decomposition, random projection, model selection criteria.


FULL TEXT

Author(s):

Ревунова Елена Георгиевна, канд. техн. наук, старший научный сотрудник Международного научно-учебного центра информационных технологий и систем НАН Украины и МОН Украины, Киев,
e-mail: egrevunova@gmail.com