Meta-Learning to Communicate: Fast End-to-End Training for Fading Channels

TL;DR

This paper introduces a meta-learning approach for end-to-end communication over fading channels, enabling rapid adaptation and training with minimal data, significantly speeding up the process compared to traditional methods.

Contribution

It proposes a meta-learning framework that finds a common initialization for fast adaptation to any fading channel, improving training efficiency over joint training methods.

Findings

Achieves significant training speed-ups with as little as one gradient step.

Demonstrates effective communication models on various fading channels.

Validates approach through numerical experiments showing rapid convergence.

Abstract

When a channel model is available, learning how to communicate on fading noisy channels can be formulated as the (unsupervised) training of an autoencoder consisting of the cascade of encoder, channel, and decoder. An important limitation of the approach is that training should be generally carried out from scratch for each new channel. To cope with this problem, prior works considered joint training over multiple channels with the aim of finding a single pair of encoder and decoder that works well on a class of channels. As a result, joint training ideally mimics the operation of non-coherent transmission schemes. In this paper, we propose to obviate the limitations of joint training via meta-learning: Rather than training a common model for all channels, meta-learning finds a common initialization vector that enables fast training on any channel. The approach is validated via numerical results, demonstrating significant training speed-ups, with effective encoders and decoders obtained with as little as one iteration of Stochastic Gradient Descent.