Broken-motifs diffusion LMS algorithm for reducing communication load

• Published in: Signal processing Vol. 133; pp. 213 - 218
• Main Authors: Chen, Feng, Shao, Xiaodan
• Format: Journal Article
• Language: English
• Published: Elsevier B.V 01.04.2017
• Subjects: Communication reduction; Diffusion LMS; Network motifs; Wireless sensor network; Wireless sensor network; Network motifs; Communication reduction; Diffusion LMS
• ISSN: 0165-1684; 1872-7557
• DOI: 10.1016/j.sigpro.2016.11.005

This paper considers the diffusion LMS (DLMS) algorithm over wireless sensor network with motifs. Network motifs are local structural patterns which are quite common in wireless sensor network. In this paper, it is found that the estimation performance of DLMS algorithm has regular changes with the varying of motifs. This result supplies some guidelines on how to choose network with optimized estimation performance. Accordingly, a broken-motifs diffusion LMS algorithm (BM-DLMS) is proposed in which only a subset of edges are participated in communications per iteration. Simulations present that the BM-DLMS algorithm efficiently reduces the communication load with less performance degradation compared with other selection algorithms. •It is found that the fewer motifs with closed triangles in the same scale networks, the better the diffusion estimation performance.•A broken-motifs diffusion LMS algorithm (BM-DLMS) is proposed in which only a subset of edges are participated in communications per iteration.•Simulations present that the BM-DLMS algorithm efficiently reduces the communication load with less performance degradation compared with other selection algorithms.