Generative Adversarial Network-based Channel Modeling for Free-Space Optical Communication

Two deep learning-based differentiable models based on GAN and BiLSTM are proposed for FSO channel modeling. Through contrastive analysis, it shows that GAN is more suitable for learning FSO channel distribution with KL divergence<0.55.