Intelligent Optical Communication Based on Wasserstein Generative Adversarial Network

Fig. 1. End-to-end communication system. (a) Structure diagram of end-to-end communication system; (b) structure diagram of automatic encoder and automatic decoder

Download full size

Fig. 2. Diagram of GAN structure

Download full size

Fig. 3. Training structure of receiver, transmitter, generator

Download full size

Fig. 4. BLER curve in AWGN channel

Download full size

Fig. 5. BLER curve in Lognormal channel

Download full size

Fig. 6. Change of accuracy and loss value with number of samples. (a) Loss value; (b) accuracy

Download full size

Table 1. Model parameter

View table

Table 1. Model parameter

Parameter	Value
Hidden layer of transmitter	32, 32
Transmitter learning rate	0.0005
Hidden layer of receiver	32, 32
Receiver learning rate	0.0005
Hidden layer of generator	128, 128, 128
Hidden layer of discriminator	32, 32, 32
Generator and discriminatorlearning rate	0.0001

Table 2. Lognormal channel parameters
View table
Table 2. Lognormal channel parameters
Parameter Symbol Value
Link distance L /km 100
Laser wavelength λ /nm 1550
Photoelectric conversion efficiency R 1
Receiver diameter D /mm 200
Refraction parameter $C_{n}^{2}$ 2.7×10^-18
Rytov variance σ_R 0.24

Tools

Get Citation

Copy Citation Text

Mu Di, Meng Wen, Zhao Shanghong, Wang Xiang, Liu Wenya. Intelligent Optical Communication Based on Wasserstein Generative Adversarial Network[J]. Chinese Journal of Lasers, 2020, 47(11): 1106005

Download Citation

EndNote(RIS)BibTex Plain Text

Set citation alerts for article

Save article for my favorites