The paper analyzes current research and the state of the industry to assess the complexity of machine learning algorithms. The tasks of deep learning are associated with an extremely high degree of computational complexity, which requires the use, first of all, of new algorithmic methods and an understanding of the assessment of the complexity of the calculations. This area of research is not given due attention for various reasons, but primarily because of the novelty of this paradigm, as well as the use of other advanced methods, which is briefly analyzed in this paper.