Jacobi Iteration based Distributed Regression in Wireless Sensor Networks
Chaojun Hou, Guoli Wang · INTERNATIONAL JOURNAL ON Advances in Information Sciences and Service Sciences · 2012
Data modeling is a potential way of leveraging data spatio-temporal correlations for energy-saving data acquisition in wireless sensor networks (WSNs). This article presents energy-efficient Jacobi distributed algorithm for solving the algebraic equation towards to data regression modeling. In particular, the distributed iterative scheme is proposed to seek the solution of regression algebraic equation with the mixture representation of the regression model. In doing this, an elegant tree-based routing structure is employed to organize the nodes in coordination to accomplish the in-network implementation of the distributed iterative scheme. Under this routing scheme, the intersections among nodes are decoupled via asynchronous message passing protocol, which ensure each node can receive the exact decoupled information from local neighboring nodes for updating its local models. Rather than transmitting the raw sensed data to base station through the network, our Jacobi distributed algorithm provide an energy-saving data acquisition by only collecting the optimal weighted coefficients. The experimental results show that the method proposed here outperforms over the existing schemes in both computation costs and transmission requirements.