Industrial Automatic Control Systems and Controllers Annotation << Back Train Machine Learning Models Using Modern Containerization and Cloud Infrastructure V.V. Nikonov, A.V. Gorchakov The article proposes a distributed system for training machine learning models, which allows renting cloud servers on Microsoft Azure services, launching computations in containerized environments, and canceling lease of servers upon completion of computations. to test an information system that automates the process of running distributed algorithms in the Microsoft Azure cloud. The application of the hybrid FSSGD algorithm based on the evolutionary method of fish school search and finite-difference gradient descent in machine learning is considered. Keywords: intelligent systems; decision support systems; expert systems; predictive systems; data analysis; machine learning; containerization; cloud infrastructure; evolutionary algorithms; distributed system; optimization algorithms; server architecture. DOI: 10.25791/asu.6.2021.1288 Pp. 26-36.