Combinatorial optimization theory and artificial intelligence problems

Traditional approaches to the analysis of operations, techniques for choosing the most effective solution are the main tools in the theory of combinatorial optimization. However, with the problem scale increase, the accumulation of significant information amounts in practical applications, traditional algorithms face great computational pressure. The solution space of combinatorial optimization problems tends to grow exponentially with increasing problem size. Therefore, an important direction for simplifying and obtaining a more accurate solution for this class of problems is the use of artificial intelligence technologies that allow us to overcome the problem of too large an action space. In the research presented in the article, the possibilities of combinatorial optimization based on a hybrid pipeline «machine learning — combinatorial optimization» based on a graph neural network, which allows to implement algorithmic alignment, are studied. Keywords optimization, algorithm, neural network, problem, dimension, space, modeling