Nonlinear feedback codes improve communication over noisy channels
Nonlinear Posterior-Mean Feedback Codes for AWGN Channels
Information Theory
Summary
Sending information over noisy channels can be made more reliable by using feedback from the receiver. The authors propose a new way to design communication codes that use feedback in a nonlinear way, making the process easier to understand and more efficient. Their method works well even when the feedback itself is noisy and uses fewer parameters than other advanced approaches. Tests show these codes perform better than existing techniques for certain message sizes.
What this means in practice
- •For wireless communication engineers: Design communication protocols that maintain message accuracy over noisy wireless channels using interpretable nonlinear feedback codes with fewer parameters.
- •For satellite communication teams: Improve the reliability of data transmissions in satellite links where feedback is available but can be noisy, by using these new feedback coding schemes.
Authors
Yingyao Zhou, Natasha Devroye, Milos Zefran, Gyorgy Turan
Abstract
Feedback can improve the reliability of communication over additive white Gaussian noise (AWGN) channels. Classical feedback codes are interpretable but often rely on linear estimation, while deep-learned feedback codes can achieve strong performance but require many learned parameters and are difficult to interpret. In this work, we propose an interpretable posterior-mean feedback coding framework for AWGN channels with feedback. The proposed scheme uses posterior-mean refinement to construct nonlinear feedback codes under both noiseless passive feedback and noisy active feedback, with maximum a posteriori (MAP) decoding at the receiver. We further develop a projection-based design to improve robustness under noisy feedback and support larger message sizes. Numerical results show that the proposed schemes achieve strong finite-blocklength performance and outperform several analytical and learned feedback coding baselines, while using only a small number of learned design parameters.