The algorithm of artificial immune system simulation with Saaty selection operator and one-dimensional local search

Zheldak, T. A.; Slesarev, V. V.; Gulina, I. G.

Purpose. Development of an algorithm which implements a certain method of modelling an artificial immune system for solving the task of multidimensional constrained optimization of multiextremum continuous functions and which provides increasing performance indicators. Methodology. A hybrid adaptive immune algorithm is proposed. It uses an operator of clonal selection based on the evaluation of fitness of solutions using the method of Saaty's hierarchy; pair adaptive crossover; the adaptive mutation based on polynomial and normal distribution laws; limited coordinate wise local search using the method the Golden section. Findings. The use of the algorithm for multidimensional constrained optimization that simulates the behavior of the artificial immune system as the operator of selection is mathematically justified by the operator based on the method of analysis of Saaty's hierarchies. It is proposed for the first time. Unlike any of known implementations of heuristic operators, it ensures a high precision and speed of the ascent up to the same level of problems. Originality. The results show high efficiency of the proposed algorithm for optimization of standard objective functions used as a test on the spacial dimensions up to 100 iterations. The algorithm shows stable convergence and a higher speed when there is a task of training neural networks of direct distribution. Practical value. The main advantages of the proposed algorithm lie in the fact that it remains effective with the growth of the of the problem; it finds not one solution, but many of them (alternatives); use much less time for a comparable solution. These properties allow the application of the proposed algorithm to solve multi-criteria multi-factor optimization problems of decision making in the processes of complex systems control.


	Наукова періодика України		Naukovyi visnyk Natsionalnoho Hirnychoho Universytetu