Finding network modules by ranking pathways

Computer networks, social networks, food webs, protein interactions, genetic regulatory networks, and many other distributed computational systems share not just the underlying structures, but also the modes of information processing. Moreover, they share the problems of scope and scaling. In order to understand, control, and sometimes secure the correct functioning of complex networks, we need to decompose them into smaller, more manageable parts. One aspect of this task is addressed in the large body of work on graph partition, clustering and classification, which seeks to identify those network nodes which play similar roles. A more recent effort has been directed towards developing the methods to recognize and structure the functional network subunits, spanned by the nodes which act together. Such functional subunits are often called modules. In this talk, i shall briefly survey the problem of network modularity, and argue that it requires an analysis of network information flows, and not just of the link structure. As a step in this direction, I shall present a method for extracting modules from a network, and for recognizing their own network structure, based on the idea of ranking the pathways that carry the information flows.

In this talk, i shall briefly survey the problem of network modularity, and argue that it requires an analysis of network information flows, and not just of the link structure. As a step in this direction, I shall present a method for extracting modules from a network, and for recognizing their own network structure, based on the idea of ranking the pathways that carry the information flows.