Deep learning-based optical vector-eigenmode processing

Optical vector-eigenmode processing for no matter few-mode fibre- or free-space-based systems is crucial to the optical information processing (OIP). In this paper, we review our latest research on the application of deep-learning in laser communications. Firstly, we introduce the deep-learning approach into the vector-eigenmode decomposition of few-mode-based transmission systems, and investigate the decomposition performance through ResNet in detail. Subsequently, we explore the laser communication in free space, where the atmospheric turbulence causes the distortion of the received beam. We employ the generative adversarial network (GAN) to achieve the image restoration, resulting into an accurate reconstruction outcome.