main| new issue| archive| editorial board| for the authors| publishing house|
Main page
New issue
Archive of articles
Editorial board
For the authors
Publishing house



No. 12. Vol. 25. 2019

DOI: 10.17587/it.25.707-717

T. A. Agasiev, Ph.D. Student, Assistant, e-mail:, Bauman Moscow State Technical University, Moscow, 105005, Russian Federation

Sectorization Method of Full Variability Map Analysis for Characteristic Features Assessment of Global Optimization Problems

Landscape analysis methods are designed to assess various characteristic features of objective function of optimization problem. The accuracy of estimations mostly depends on the chosen design of experiment, i.e. on the number and location of points in the search space. The paper considers methods of variability map analysis that allow to obtain estimations that are more stable to variations of the experimental design. The drawback of these methods is the high sensitivity of resulting values of characteristic features to the scale of objective function values variations. This may adversely affect the generalizing ability of the algorithm of problems classification by the most significant objective function features. In relation to this, the paper presents sectorization method of variability map analysis which allows to obtain estimations stable to the scale of objective function changes. A technique of conducting the computational experiment is proposed for comparative research of the efficiency of different methods of variability map analysis. Efficiency criteria for algorithms of characteristic features assessment are formulated. The results of experiment demonstrate relevance of using the proposed methods of landscape analysis for classifying optimization problems by different features. The quality of optimization problems classification largely defines the efficiency of intellectualization of the parametric optimization subsystems in CAD.
Keywords: global optimization, characteristic features of problem, exploratory landscape analysis, landscape sample, information content method, sectorization method, variability map

P. 707Ц717

To the contents