This work is devoted to the study of the classification accuracy of optical vortex superposition parameters using convolutional neural networks. Much of the work is devoted to analyzing ways to improve classification accuracy. The paper presents the results of experiments on training convolutional neural networks on different training and test datasets. The fact of a significant advantage of using one neural network architecture over another in the context of the problem being solved has been established.
Shilov et al. (Mon,) studied this question.