Skip to Main Content
This article explained how nodes in a network graph can infer information about the network topology or its topology related properties, based on in-network distributed learning, i.e., without relying on an external observer who has a complete overview over the network. Some key concepts from the field of SGT were reviewed, with a focus on those that allow for a simple distributed implementation, i.e., eigenvector or Katz centrality, algebraic connectivity, and the Fiedler vector. This paper also explained how the nodes themselves can quantify their individual network-wide influence, as well as identify densely connected node clusters and the sparse bridge links between them. The addressed concepts, as well as more advanced concepts from the field of SGT, are believed to be crucial catalysts in the design of topology-aware distributed algorithms. Examples were provided on how these techniques can be exploited in several nontrivial distributed signal processing tasks.