CAREER: Graph Identification

University Of Maryland, College Park, College Park MD

Investigators

Abstract

Today's world requires that we participate in, comprehend, or otherwise navigate many kinds of networks, including social networks, communication networks, financial transaction networks, gene regulatory networks, disease transmission networks, ecological food networks, sensor networks and more. Observational data on such networks is plentiful, but extracting the relevant data is challenging. Furthermore, this data often has errors, is incomplete and fragmented, or otherwise does not offer a concise and big-picture view of the network that we want to understand and use. This research project will develop methods for inferring and representing the essential, underlying network structure that best accounts for observed entities and interactions between these entities. In particular, the proposal introduces a process abstraction called ?graph identification,? which is composed of more basic processes for identifying observed entities that are equivalent, identifying links (i.e., relationships) between entities from observed interactions, and simultaneously classifying many entities into classes based on their interconnection structure. A computational challenge of this project is integrating these basic processes, each of which has been studied by the PI and others, to achieve accurate solutions to the overall graph identification problem, in a manner that is efficient and scalable. This research has broader impacts for basic science, national security and a better understanding of privacy issues in today's networked society.