Online Meta-Learning Channel Autoencoders for Dynamic Wireless Fading Channels
A channel-autoencoder framework (OML-CAE) that meta-trains online as new fading-channel realizations arrive, treating each channel as a meta-learning task whose few pilot signals serve as the support and query sets. The receiver adapts to a new Rayleigh fading channel from only a few pilots, cutting pilot overhead roughly 3.8x compared with retraining a conventional channel autoencoder while achieving lower symbol error rates. Students learn how MAML-style inner and outer loops map onto the transmitter-receiver adaptation problem and why few-shot online adaptation matters for real-time physical-layer design.
Online Meta-Learning Channel Autoencoder for Dynamic End-to-end Physical Layer Optimization
Channel Autoencoders (CAEs) have shown significant potential in optimizing the physical layer of a wireless communication system for a specific channel through joint end-to-end training. However, the…