Cooperation gain in steady-state performance of diffusion adaptive networks with noisy links

Recently, adaptive networks have been widely considered as a powerful solution to distributed information processing problems. In this paper, we provide some insights into the steady-state performance of diffusion least-mean-squares (D-LMS) adaptive network with noisy links. To this end, we define the concept of cooperation gain in adaptive networks. Then, we show that when the connecting links are ideal (error-free), the cooperation between nodes in adaptive networks always leads to a smaller steady-state error than what can be achieved by a non-cooperative scheme. While, with noisy links, cooperation between the nodes does not always provide better result, only in specific conditions leads to better results. We demonstrate the performance of the system by performing a number of simulations.