For example,Бобцов

ADAPTIVE SELECTION OF ARTIFICIAL IMMUNE SYSTEMS AND LOCAL SEARCH MUTATION OPERATORS

Annotation

Subject of Research.Evolutionary algorithms use various mutation operations, which can be optimal at different stages of optimization. We formulate the task of choosing the most suitable mutation operator while optimizing. We propose a method for adaptive selection of mutation operators with probability depending on the current fitness. This property makes this method efficient on every stage of optimization. Method. We use two classes of algorithms: artificial immune systems, which are efficient at the initial stage of optimization, and randomized local search, which is efficient towards the end. The new method and the existing algorithms are compared experimentally on two benchmark problems. Main Results. The method for adaptive selection between artificial immune systems and local search mutation operators is developed. An experimental comparison of the proposed method with existing ones was performed. It showed that the proposed method is efficient under various computational budgets and finds optimal problem solutions faster than the other methods. Practical Relevance. The proposed modification improves the performance of algorithms when solving dynamic optimization problems with fitness functions changing in time, such as Pickup-and-Delivery Problem.

Keywords

Постоянный URL

Articles in current issue

CONTENT © ITMO UNIVERSITY 1993-2024