Gossiping in Groups: Distributed Averaging over the Wireless Medium
The authors present an approach to gossip algorithms tailored to the practical considerations of wireless communications. Traditional gossip algorithms operate via the pair-wise exchange of estimates, which fails to capture the broadcast and superposition nature of the wireless medium. Adapting the virtual full-duplex framework of Guo and Zhang, they construct a communications scheme in which each node can broadcast its estimate to its neighbors while simultaneously receiving its neighbors' estimates. This full-duplex scheme gives rise to group gossip, a more flexible family of gossip algorithms built on multilateral, rather than pair-wise, exchanges.