Learning to Broadcast with Layered Division Multiplexing

Abstract

A broadcast/multicast communication system is studied in which layered division multiplexing (LDM) is applied to support differential quality-of-service (QoS) levels. Focusing on a practical scenario in which the transmitter does not know the fading distribution, layer allocation is optimized based on a dataset sampled during deployment. The optimality gap caused by the availability of limited data is bounded via a generalization analysis, and is shown to be monotonically decreasing as the dataset grows larger. Numerical experiments demonstrate that LDM improves spectral efficiency even for small datasets; and that, for sufficiently large datasets, the proposed mirror-descentbased layer optimization scheme achieves an expected rate close to that achieved when the transmitter knows the fading distribution.