Distributed Quantization-Aware RLS Learning With Bias Compensation and Coarsely Quantized Signals

Abstract

In this work, we present an energy-efficient distributed learning framework using coarsely quantized signals for Internet of Things (IoT) networks. In particular, we develop a distributed quantization-aware recursive least-squares (DQA-RLS) algorithm that can learn parameters in an energy-efficient fashion using signals quantized with few bits while requiring a low computational cost. Moreover, we develop a bias compensation strategy to further improve the performance of the proposed DQA-RLS algorithm. We carry out a statistical analysis of the proposed DQA-RLS algorithm and derive analytical expressions for predicting the mean-square deviation. A computational complexity evaluation and a study of the power consumption of the proposed and existing techniques are then presented. Numerical results assess the DQA-RLS algorithm against existing techniques for a distributed parameter estimation task in a scenario where IoT devices operate in peer-to-peer mode.