Deep convolutional generative adversarial networks: performance analysis in wireless systems

Abstract

Abstract This paper introduces Deep Convolutional Generative Adversarial Networks (DCGAN) as a potential solution for wireless systems aiming to enhance the Block Error Rate (BLER). The DCGAN under consideration consists of a generator and a discriminator, which utilizes the same network to differentiate between true and synthetic messages while simultaneously reconstructing the transmitted message. The wireless network is modeled as a DCGAN, where the desirable communication link is modeled between the transmitter and receiver. In addition, the expression for the transmitter and receiver of DCGAN is analytically derived. Based on the simulation outcomes, the proposed DCGAN has investigated a significantly lower BLER than the other models, namely convolutional neural networks, deep learning, and conditional GAN.

Keywords

• Block error rate
• Convolution neural network
• Deep learning
• Discriminator
• Generator
• Generative adversarial network