Denoising in Mode Conversion by Utilizing Diffractive Deep Neural Networks Optimized with Reinforcement Learning
- Resource Type
- Conference
- Authors
- Li, Zheng; Zhang, Wenbo; Wang, Yang; Peng, Guanju; Li, Zongze; Zhou, Xiaoyan; Zhang, Lin
- Source
- 2024 Optical Fiber Communications Conference and Exhibition (OFC) Optical Fiber Communications Conference and Exhibition (OFC), 2024. :1-3 Mar, 2024
- Subject
- Communication, Networking and Broadcast Technologies
Photonics and Electrooptics
Gaussian noise
Noise reduction
Reinforcement learning
Artificial neural networks
Optical fiber communication
- Language
We propose a reinforcement-learning-optimized nonlinear physical diffractive neural network, which can simultaneously perform OAM-mode and LP-mode conversion with Gaussian noise removal. The PSNR and SSIM of the converted modes reach 27.94 dB and 0.838, respectively.