Multi-Task Diffusion With Masked Measurements

Abstract

This letter addresses the problem of clustered multitask distributed estimation under masked measurements, where network nodes observe partial or incomplete data due to sensing limitations, communication constraints, or privacy requirements. We propose a novel extension of the Diffusion LMS (DLMS) algorithm that incorporates node-specific masking and a task-clustered structure. A tailored network-wide optimization problem is formulated to jointly handle masked observations and inter-cluster multitask estimation. Convergence analysis and simulation results demonstrate the effectiveness and robustness of the proposed approach in improving estimation performance under partial observability.