Grantee: University of North Carolina at Chapel Hill, Chapel Hill, NC, USA
Researcher: Peter J. Mucha, Ph.D.
Grant Title: Community Detection in Networks Across Time
https://doi.org/10.37717/220020315
Program Area: Studying Complex Systems
Grant Type: Scholar Award
Amount: $450,000
Year Awarded: 2012
Duration: 6 years
Interdisciplinarily rooted in graph theory and the long development of social network analysis, the study of complex networks has exploded in the past two decades. With continued developments across the quantitative social sciences and newer emphases in statistical physics, applied mathematics, computer science, and statistics, network analytics has proven important for the study of a diverse array of complex systems, with successful and important applications across social, political, anthropological, biological, chemical, technological, epidemiological, financial, and informational processes as ever larger numbers and sizes of data become available. With nodes representing selected entities or agents and edges representing relationships or connections between nodes, networks can be used to study web pages, consumer data for recommendation systems, phone calls between individuals, mutual fund holdings, protein interactions, disease-spreading contacts, or nano-scale particles in heterogeneous materials, just to name a few examples. The potentially astoundingly large scales of such data demand new techniques for understanding and classifying these systems, utilizing and further developing a collection of best practices from across the wide variety of interdisciplinary activity ongoing in the study of networks.
One of the prominent developments in networks, particularly over the past decade, is the algorithmic detection of mesoscopic structures known as communities, intuitively conceived as groups of nodes connected more tightly to each other than to the rest of the network. Community detection has become an excellent way to characterize and understand the mesoscopic structures of a network. Such studies repeatedly harken to the connection between structure and function: the processes in network-connected systems are both influenced by and drive the structural properties of the collection of edges between nodes, so a better understanding of structure yields insights about function. The algorithmic detection of communities has thus become a highly powerful technique used successfully in the study of many diverse real-world networks.
Until recently, however, the study of community detection in networks that vary over time has been relatively limited compared to the wealth of methods and studies on static network representations. This situation changed with the principled extension of the community detection concept known as modularity (and its quality function generalizations) to networks which vary over time or which encode multiple different kinds of connections (multiplex networks) by Mucha et al. (2010). This multislice modularity methodology opens wide the door to numerous applications that were previously inaccessible to most community detection heuristics. At the same time, the modularity framework itself is well known to have a number of pitfalls, and there are many other well-known and well-studied methods for community detection of static (i.e. single-slice) networks. Motivated by the successes of the multislice modularity approach, Mucha’s research program seeks to similarly extend other state-of-the-art community detection methods to multislice network data. Such developments will greatly expand the utility of community detection, with immediate, significant, and broad impact on the study of real-world networks for applications across a wide array of disciplines.