
PresentationsResonant nature of setting the control parameters of the optimization of the differential evolution algorithmProkhorov General Physics Institute of the Russian Academy of Sciences, Vavilova street, 38, V.K27@yandex.ru, rpishchal@kapella.gpi.ru Differential evolution is one of the most widely used evolutionary optimization methods, the heuristic nature of which allows it to be applied to a wide range of tasks, without going into the nuances and details of the model under study. We have developed a software in which you can use as an optimized function any preset mathematical function. Also, narrowly focused procedures for calculating semiclassical quantum parameters necessary for the calculation of optical properties of organic pigments and crystals [13]. In this paper, we investigated the convergence of the algorithm on a set of test functions (both for unimodal and multimodal), which represent an unsolvable problem for classical optimization methods, such as gradient descent or the method of least squares. As a result, the resonant nature of the increase in the convergence rate of differential evolution was found when the two main parameters of the algorithm were varied: F (weight coefficient) and Cr (crossover verticality). This feature of the algorithm should always be kept in mind, since the nonoptimal selection of F and Cr can significantly reduce the efficiency of optimization up to a complete lack of convergence. Literature 1. Pishchalnikov R. Application of differential evolution for modeling the linear optical response of photosynthetic pigments // Journal of Computational Physics, volume 372, 2018, pp. 603615, doi:10.1016/j.jcp.2018.06.040. 2. Pishchalnikov, R.Yu.; Yaroshevich, I.A.; Zlenko, D.V.; Tsoraev, G.V.; Osipov, E.M.; Lazarenko V.A.; Parshina E.Yu.; Chesalin D.D.; Sluchanko N.N.; Maksimov E.G. The role of the local environment in the structural heterogeneity of betaion rings of carotenoids // Investigation of photosynthesis, volume 156, November 1, 2022, pp. 317, doi:10.1007/s11120022 009552. 3. Chesalin, D.D.; Kulikov, E.A.; Yaroshevich, I.A.; Maksimov, E.G.; Selishcheva, A.A.; Pishchalnikov R.Yu. Differential evolution reveals the influence of polar and nonpolar solvents on carotenoids: an example of modeling the optical response of astaxanthin // Swarm and evolutionary calculations, 75, 2022, 101210, doi:10.1016/j.swevo.2022.101210
Presentation 