All-dielectric nanophotonics is a rapidly growing field of modern science. Metasurfaces and other planar devices based on all-dielectric nanoparticles lead to manage the light propagation at the nanoscale. Impressive effects such as perfect absorption, invisibility, chirality effects, negative refraction, light focusing in the area with size smaller than wavelength, nano-lasing etc - can be achieved with all-dielectric technologies. While it is needed to use more and more complicated designs for solution of modern nanophotonics' currents tasks, non-classical methods of optimization become relevant. For example, to design reconfigurable metalenses with an additional degree of freedom such as polarizability or temperature dependence, evolutionary or genetic algorithms show their high applicability. In this work, we show a new approach to design metalenses with evolutionary and genetic algorithms. © 2020 IOP Publishing Ltd.