Learning Robust Representations for Communications over Noisy Channels

TL;DR

This paper proposes a novel neural network-based approach for designing end-to-end communication systems that are robust to noise, using information theory principles and innovative training strategies.

Contribution

It introduces a new encoder structure inspired by Barlow Twins and demonstrates the effectiveness of iterative training with varying noise levels for robust communication.

Findings

Iterative training with random noise levels improves error performance.

Mutual information and pairwise distance-based cost functions enhance robustness.

The proposed method outperforms traditional approaches under power constraints.

Abstract

We explore the use of FCNNs (Fully Connected Neural Networks) for designing end-to-end communication systems without taking any inspiration from existing classical communications models or error control coding. This work relies solely on the tools of information theory and machine learning. We investigate the impact of using various cost functions based on mutual information and pairwise distances between codewords to generate robust representations for transmission under strict power constraints. Additionally, we introduce a novel encoder structure inspired by the Barlow Twins framework. Our results show that iterative training with randomly chosen noise power levels while minimizing block error rate provides the best error performance.