Development and research of neural network algorithms for fast reconstruction of particle tracks

The paper is devoted to the development of effective methods for solving the problem of reconstruction particles tracks (trajectories of motion), generated by the collision of accelerated charged particles or atomic nuclei with a stationary target. Important properties of this problem are: the need to solve it in real time; large volumes of processed data; a significant level of noise in this data due to the features of the design of track detectors. To solve this problem, three methods are proposed: based on the Hough transform, using the technology of self-organizing Kohonen maps and graph neural networks. The results of a numerical study of the proposed methods in solving a model problem of particle tracking are presented.