Machine learning for the search for topological spin textures

We present an alternative method for numerical modeling of topological magnetic textures using a neural network algorithm. We discuss a model of localized spins where topological magnetic textures are stabilized due to a delicate interplay between the symmetric exchange interaction, and the antisymmetric interaction caused by exchange–relativistic effects, as well as a model of an itinerant magnet where noncoplanar spin configurations emerge when taking multispin interactions into account. The viability of the proposed method is illustrated with the formation of lattices of skyrmions and antiskyrmions, magnetic hedgehogs, and skyrmion tubes in two-dimensional and three-dimensional magnetic systems.