Laser & Optoelectronics Progress, Volume. 58, Issue 11, 1106001(2021)
Intelligent Optical Communication Based on Self-Organizing Incremental Generative Adversarial Network
In this article, first, the advantages of laser communication links are introduced. Then, end-to-end learning based on generative adversarial network improves the real-time and global optimization of the communication system. Finally, because of the lack of robustness in the system caused by the training set cannot contain all situations during the training process of the offline learning mode of the system, a self-organizing incremental learning method is introduced to improve online incremental training of the end-to-end system. Experiment results show that by implementing the self-organizing incremental learning method, the end-to-end communication system can simulate the channel effect efficiently, and it greatly improved the real-time optimization and robustness of the system.
Get Citation
Copy Citation Text
Di Mu, Wen Meng, Wenya Liu, Xiang Wang, Shanghong Zhao. Intelligent Optical Communication Based on Self-Organizing Incremental Generative Adversarial Network[J]. Laser & Optoelectronics Progress, 2021, 58(11): 1106001
Category: Fiber Optics and Optical Communications
Received: Sep. 12, 2020
Accepted: Oct. 29, 2020
Published Online: Jun. 7, 2021
The Author Email: Meng Wen (meng-wen_mw@126.com)