Abstract:The practical realization of end-to-end training of communication systems is fundamentally limited by its accessibility of the channel gradient. To overcome this major burden, the idea of generative adversarial networks (GANs) that learn to mimic the actual channel behavior has been recently proposed in the literature. Contrarily to handcrafted classical channel modeling, which can never fully capture the real world, GANs promise, in principle, the ability to learn any physical impairment, enabled by the data-driven learning algorithm. In this work, we verify the concept of GAN-based autoencoder training in actual over-the-air (OTA) measurements. To improve training stability, we first extend the concept to conditional Wasserstein GANs and embed it into a state-of-the-art autoencoder-architecture, including bit-wise estimates and an outer channel code. Further, in the same framework, we compare the existing three different training approaches: model-based pre-training with receiver finetuning, reinforcement learning (RL) and GAN-based channel modeling. For this, we show advantages and limitations of GAN-based end-to-end training. In particular, for non-linear effects, it turns out that learning the whole exploration space becomes prohibitively complex. Finally, we show that the training strategy benefits from a simpler (training) data acquisition when compared to RL-based training, which requires continuous transmitter weight updates. This becomes an important practical bottleneck due to limited bandwidth and latency between transmitter and training algorithm that may even operate at physically different locations.

Unknown Channel End-to-End Learning of Communication System With Residual DCGAN.

Residual-Aided End-to-End Learning of Communication System without Known Channel

A Learning-Based End-to-End Wireless Communication System Utilizing a Deep Neural Network Channel Module

Channel Agnostic End-to-End Learning based Communication Systems with Conditional GAN

ChannelGAN: Deep Learning-Based Channel Modeling and Generating

Ensemble-Based Distributed Learning for Generative Adversarial Networks.

DCC-GAN-based channel emulator for underwater wireless optical communication systems

Deep reinforcement learning autoencoder with RA-GAN and GAN

Learning End-to-End Channel Coding with Diffusion Models

DCCRGAN: Deep Complex Convolution Recurrent Generator Adversarial Network for Speech Enhancement

Deep Learning based Modeling of Wireless Communication Channel with Fading

Attention-Empowered Residual Autoencoder for End-to-End Communication Systems

Channel model for end-to-end learning of communications systems: A survey

Generative Adversarial Network for Wireless Communication: Principle, Application, and Trends

WGAN-based Autoencoder Training Over-the-air

Deep Learning for Channel Coding via Neural Mutual Information Estimation

An Effective Generative Model Based Channel Estimation Method With Reduced Overhead

Channel Attention Image Steganography With Generative Adversarial Networks

A Multi-Task Learning Model for Super Resolution of Wireless Channel Characteristics

A Robust Adversarial Network-Based End-to-End Communications System With Strong Generalization Ability Against Adversarial Attacks

Learning Robust Representations for Communications over Noisy Channels