АДАПТАЦІЯ ЛОКАЛЬНО-ДЕТЕРМІНІСТИЧНИХ МЕТОДІВ ОПТИМІЗАЦІЇ ДО РОЗВ’ЯЗАННЯ ЗАДАЧІ ОЦІНКИ ПАРАМЕТРІВ МАТЕМАТИЧНОЇ МОДЕЛІ МЕТОДОМ ГАУСА-МАРКОВА

Abstract

This article contains attempt to find the most important traits of parameter estimation problem and on basis of them to propose demands’ list for optimizers that solve this problem. The most suitable groups of locally deterministic optimization methods for solving parameter estimation problems are outlined and corresponding modifications for them are suggested and considered. The special attention is given to quasi-Newton optimization methods’ research. Pipelining preprocessor hybrid genetic algorithms is used for search of global minimum. As a result of its research new conversion criterion was proposed.