A novel adaptive algorithm for diffusion networks using projections onto hyperslabs

Symeon Chouvardas, Konstantinos Slavakis, Sergios Theodoridis

Abstract: In this paper, a new algorithm for distributed learning in sensor networks is developed. The algorithm is built upon a diffusion protocol to implement cooperation among neighbouring nodes. The algorithm is developed in the convex set theoretic approach, and it is based on a sequence of metric projections on hyperslabs. Full convergence results have been obtained and the experimental set up demonstrates significant performance improvements, compared to previously derived algorithms of similar complexity.