Determination of the Parameters of the Laguerre-Gaussian Beams with an Increased Dark Region after Propagation in Turbulent Media

Abstract This paper is devoted to the study of the classification of parameters of Laguerre-Gaussian optical vortices with an increased dark region using convolutional neural networks. A significant part of the work is dedicated to the analysis of the obtained classification results. The ResNet architecture was shown to be more suitable for this task than the alternatives, achieving high classification accuracy of 99.35% for radial index (n) and 98.25% for azimutal index (m). In contrast, AlexNet demonstrated less impressive results with 98.60 and 97.45% for indeces n and m, respectively. The VGG architecture, on the other hand, performed less effectively, with accuracies of only 61.90% for parameter n and 60.75% for parameter m, rendering it unsuitable for this task.