Adaptation and Learning over Networks

Complex networks are very popular in modern science. In recent years, several research efforts to decipher the intricacies of complex networks have been progressing almost independently across several disciplines, including signal processing, machine learning, optimization, control, statistics, physics, biology, economics, computer science, and social sciences. In all these fields, there is growing interest in performing inference and learning over graphs, deducing relationships from connections over social networks, modeling interactions among agents in biological networks, diffusing information among distributed agents, optimizing functions defined over graphs, etc. In particular, we are interested in designing learning algorithms for adaptive networks, which are composed of a set of nodes, equipped with local processing and communication units, whose aim is to collectively estimate some vector parameter of interest from noisy measurements by relying solely on in-network processing. In such implementations, the nodes exchange information locally and cooperate with each other without the need for a central processor. In this way, information is processed on the fly by all nodes and the data diffuse across the network by means of a real-time sharing mechanism. The resulting adaptive networks fully exploit the time and spatial-diversity of the data, thus endowing networks with powerful learning and tracking abilities. Specific applications of the proposed methods include sparsity-aware distributed online strategies, online sampling and distributed recovery algorithms, and distributed methods for the estimation and control of the algebraic connectivity of random graphs.

Selected papers: