The development and advances of machine learning methods in a wide range of tasks significantly affected the design and implementation of scientific research in solar physics. Large data sets became an individual value in which expert efforts and significant technological resources have been invested. The research itself is now of an interdisciplinary nature and concentrated around advanced computing centers It is now possible to set large-scale problems where yesterday there was no formal mathematical formulation. This review presents the main ideas on which modern machine learning models are based, databases prepared for machine learning tasks and tools for efficient data processing. The main part of this review is the discussion of models proposed in the context of specific solar physics problems and their generalizations to other applications.