Study of Sparsity-Aware Distributed Conjugate Gradient Algorithms for Sensor Networks

TL;DR

This paper introduces sparsity-aware distributed conjugate gradient algorithms for sensor networks, improving parameter estimation performance by incorporating $l_{1}$ and log-sum penalties, and demonstrating superior convergence and accuracy.

Contribution

It develops novel sparsity-aware diffusion distributed CG algorithms using $l_{1}$ and log-sum penalties, enhancing performance over existing methods.

Findings

Improved mean square deviation (MSD) and convergence performance.

Performance close to diffusion distributed recursive least squares (Consensus-RLS).

Algorithms are reliable and applicable in various scenarios.

Abstract

This paper proposes distributed adaptive algorithms based on the conjugate gradient (CG) method and the diffusion strategy for parameter estimation over sensor networks. We present sparsity-aware conventional and modified distributed CG algorithms using $l_{1}$ and log-sum penalty functions. The proposed sparsity-aware diffusion distributed CG algorithms have an improved performance in terms of mean square deviation (MSD) and convergence as compared with the consensus least-mean square (Diffusion-LMS) algorithm, the diffusion CG algorithms and a close performance to the diffusion distributed recursive least squares (Consensus-RLS) algorithm. Numerical results show that the proposed algorithms are reliable and can be applied in several scenarios.