Semi-Supervised Variational Inference over Nonlinear Channels

TL;DR

This paper introduces semi-supervised variational inference techniques, including variational autoencoders, for decoding unknown nonlinear channels in communication systems, effectively utilizing limited pilot data and outperforming meta learning in certain conditions.

Contribution

It proposes a novel semi-supervised learning framework based on variational inference for nonlinear channel decoding, demonstrating improved performance over existing methods.

Findings

Variational autoencoders outperform other semi-supervised methods.

The approach achieves lower error rates with sufficient payload symbols.

It outperforms meta learning when enough payload data is available.

Abstract

Deep learning methods for communications over unknown nonlinear channels have attracted considerable interest recently. In this paper, we consider semi-supervised learning methods, which are based on variational inference, for decoding unknown nonlinear channels. These methods, which include Monte Carlo expectation maximization and a variational autoencoder, make efficient use of few pilot symbols and the payload data. The best semi-supervised learning results are achieved with a variational autoencoder. For sufficiently many payload symbols, the variational autoencoder also has lower error rate compared to meta learning that uses the pilot data of the present as well as previous transmission blocks.