Cluster-based Jacobi iteration for distributed regression in wireless sensor networks

This paper presents a Jacobi iterative based computational paradigm for solving the data regression in wireless sensor networks (WSNs). The in-network computational scheme is proposed to construct a mixture regression model through the cluster-based Jacobi distributed iteration, where the intersections among mixture structure of regression model are decoupled through a new cluster-based message passing protocol in an energy-efficient fashion. The cluster-based computational scheme proposed here contributes not only to easing network topology management, but also to speeding the convergent rate of distributed computation. Experimental results are reported to illustrate the validation of the proposed approach.