On Deep Learning-Based Channel Decoding

TL;DR

This paper explores deep neural networks for one-shot decoding of structured codes like polar codes, demonstrating their ability to learn decoding algorithms and generalize to unseen codewords, especially for structured codes.

Contribution

It shows neural networks can learn decoding algorithms for structured codes and introduces the normalized validation error metric to assess performance and complexity.

Findings

Neural networks achieve MAP BER performance for short codes.

Structured codes are easier for neural networks to learn.

Networks generalize to unseen codewords for structured codes.

Abstract

We revisit the idea of using deep neural networks for one-shot decoding of random and structured codes, such as polar codes. Although it is possible to achieve maximum a posteriori (MAP) bit error rate (BER) performance for both code families and for short codeword lengths, we observe that (i) structured codes are easier to learn and (ii) the neural network is able to generalize to codewords that it has never seen during training for structured, but not for random codes. These results provide some evidence that neural networks can learn a form of decoding algorithm, rather than only a simple classifier. We introduce the metric normalized validation error (NVE) in order to further investigate the potential and limitations of deep learning-based decoding with respect to performance and complexity.