Diffusion adaptation over networks with kernel least-mean-square

Distributed learning over networks has become an active topic of research in the last decade. Adaptive networks are suitable for decentralized inference tasks, e.g., to monitor complex natural phenomena or infrastructure. Most of works focus on distributed estimation methods of linear regression models. However, there are many important applications that deal with nonlinear parametric models to be fitted, in a collaborative manner. In this paper, we derive functional diffusion strategies in reproducing kernel Hilbert spaces.