Diffusion-based EM Gradient Algorithm for Density Estimation in Sensor Networks

Abstract

This paper considers mixture density estimation in an asynchronous sensor networks in a distributed manner. In the statistical literature, the maximum likelihood (ML) estimate of mixture distributions can be computed via a straightforward application of the expectation and maximization (EM) algorithm. In a random sensor networks, data are required to collected and processed at local decentralized processing units. Reformulations of standard EM-type algorithms are necessary to accommodate the characteristics of sensor networks. Existing works on the distributed EM implementation focus mainly on synchronous network. Here, we address the issue of asynchronous behaviors by proposing a diffusion-based EM gradient algorithm that updates estimates under ATC diffusion strategy. Simulation results show the robustness and scalability of the proposed approach in the presence of additional randomness of asynchronous events.