


DOI: 10.17587/it.24.370386 T. A. Agasiev, Ph.D. student, Assistant, email: agtaleh@mail.ru, A. P. Karpenko, Doctor of Sc., Professor, email: apkarpenko@mail.ru Bauman Moscow State Technical University, Moscow, 105005, Russian Federation Modern Techniques of Global Optimization. Review A global constrained optimization problem with high computational complexity of an objective function was formulated. The following terms have been defined: optimization problem features, base problem and base algorithm, metaproblem and metaalgorithm, base problem strategy, indicator of the strategy efficiency. Each optimization problem can be represented by a vector of its features, describing those properties of the problem and objective function that are meaningful in terms of optimization algorithms efficiency. The vector of problems features, estimated by means of exploratory landscape analysis methods, can be used to classify optimization problems or to gain important information about the problem definition in order to find an appropriate strategy of optimization algorithm to be used. To compare candidate strategies and select arguably the most suitable one the efficiency indicator should be determined. The article considers various ways of the estimation of optimization algorithms efficiency. The problem of finding the best algorithm's strategies based on a vector of problem's features and a given efficiency indicator is called metaoptimization problem. Statements of multiindicator, multiclass and multibudget metaoptimization problems are provided. The result of solving the metaoptimization problem is a set of the most appropriate strategies, i.e. the optimization algorithm tuned for solving particular types of optimization problems. Different approaches to automated tuning of optimization algorithms are described. A review of advanced exploratory landscape analysis and metaoptimization methods was performed based on approximately 100 articles. The article concludes with considering a modern software for solving the base and metaoptimization problems. P. 370–386 