•  

Using neural network approximation of the objective function in global optimization problems

TitleUsing neural network approximation of the objective function in global optimization problems
AuthorsS. N. Karpenko1, I. G. Lebedev1, D. V. Nadumin1
1N. I. Lobachevsky State University of Nizhny Novgorod
AnnotationThe paper considers the problem of finding the global minimum of a multiextremal function. To solve this problem, the simulation annealing algorithm was used. To speed up the process of finding the minimum, the approximation of the objective function by a neural network is applied. The results of algorithm that combines the simulation annealing algorithm using a neural network to approximate objective function and the Nelder-Meade method to refine the solution of the optimization problem are presented. The paper presentS the results of the experiments carried out with a series of multiextremal test problems. The results show a reduction in the number of calculations of the objective function while maintaining high search accuracy.
Keywordsoptimization methods, global optimization, neural networks, machine learning.
CitationKarpenko S. N., Lebedev I. G., Nadumin D. V. ''Using neural network approximation of the objective function in global optimization problems'' [Electronic resource]. Proceedings of the International Scientific Youth School-Seminar "Mathematical Modeling, Numerical Methods and Software complexes" named after E.V. Voskresensky (Saransk, July 14-18, 2022). Saransk: SVMO Publ, 2022. - pp. 111-119. Available at: https://conf.svmo.ru/files/2022/papers/paper17.pdf. - Date of access: 30.11.2022.