A novel scheme for diffusion networks with least-squares adaptive combiners

Abstract

In this paper, we propose a novel diffusion scheme for adaptive networks, where each node preserves a pure local estimate of the unknown parameter vector and combines this estimate with other estimates received from neighboring nodes. The combination weights are adapted to minimize a local least-squares cost function. Simulations carried out in stationary and nonstationary scenarios show that the proposed scheme can outperform other existing schemes for diffusion networks with adaptive combiners in terms of tracking capability and convergence rate when the network nodes use different step sizes.