Community structure in complex networks
Many complex networks exhibit some modular structure: they tend to have clusters of nodes that have many links inside clusters and few links across clusters, which are commonly called “communities”. In this presentation, I will sketch a broad overview of the topic. Two core problems of community structure will be highlighted and adressed: the problem of the so-called resolution limit and the problem of community structure in random networks. Only few methods do not suffer from the resolution limit and I will introduce one such method. The second problem will be addressed from the viewpoint of the significance of community structure. Interestingly, both problems seem to be incongruent, so that no method can address both problems simultaneously. I will conclude with some practical pointers for uncovering community structure and interpreting the results.
1. Traag, V. A., Krings, G. & Van Dooren, P. Significant scales in community detection. Scientific Reports 3, 2930 (2013). http://www.nature.com/srep/2013/131014/srep02930/full/srep02930.html
2.Traag, V. A., Van Dooren, P. & Nesterov, Y. Narrow scope for resolution-limit-free community detection. Physical Review E 84, 016114 (2011). http://link.aps.org/doi/10.1103/PhysRevE.84.016114
Vincent Traag has completed his PhD at the Department of Applied Mathematics at the Université catholique de Louvain in Belgium, after which he will join the e-Humanities Group and the KITLV. In general, he is interested in social networks, social dynamics and conflict, with a desire to combine mathematical modeling and social sciences.
Originally Traag began studying computer science, but after two years he decided to switch to sociology. However, he found the more formal analysis lacking in sociology and he took up an additional year of mathematics. After graduating cum laude in sociology at the University of Amsterdam he joined the research group on “Large Graphs and Networks” at the Department of Applied Mathematics at the Université catholique de Louvain. In his thesis he addresses two topics in network science. The first topic focuses on finding communities—groups of densely connected nodes—in networks. Some methods for finding communities suffer from a drawback: they cannot detect small communities in large graph. This problem, known as the resolution-limit, was analyzed in depth by him, and he showed that only few methods do not suffer from the problem. Secondly, he investigated negative links in networks—links that represent conflict or hatred. On the one hand, the problem is similar to community detection: the focus is on finding groups with positive links within groups, but negative links in between groups. On the other hand, he analyzed how such a structure of negative and positive links might come about through social dynamics.
2. Frank van der Most, DANS and eHumanities Group
Adding and finding meaning in case-by-case network-graphs of interviews
An explorative experiment in the combination and visualization of relational data and interview-transcription coding.
For the ACUMEN project, I collected career data from and conducted interviews with about 40 university-based researchers and 10 deans, department heads and human resources managers. Career data typically comes in the form of CVs, which are suitable for storing and coding in relational databases. Doing interviews results in notes, transcriptions and coding added to the transcriptions. This is typically done with coding software such as NVIVO, Atlas or TAMSAnalyzer. Database software usually does not produce network graphs. Coding software is good at producing network graphs, but bad at dealing with relational data. The problem then is how to combine the two and for ACUMEN I explore a few possibilities. I will present one of these and evaluate its use as a tool for exploration and analysis.
Frank van der Most started his work on the ACUMEN project in the summer of 2011 at the e-Humanities Group. His research interests are research practices, the funding and organization of research, research policies and the interactions between these three. He studied Computer Science at the University of Twente and Sciences and Arts at Maastricht University. From 1997 until 2005 he was involved in research projects in the history of technology, the policy and scientific developments surrounding mad cow disease, and an evaluation of the Norwegian Research Council. During these projects he developed an interest in digital tools for qualitative and historical research, for which he developed a database application. After a failed attempt to commercially exploit these interests he returned to academia and in 2009 defended his doctoral thesis titled ‘Research councils facing new science and technology : The case of nanotechnology in Finland, the Netherlands, Norway and Switzerland’ at the University of Twente. From 2009 until 2011, he did a post-doctoral project on the use and effects of research evaluations at the CIRCLE institute for innovation studies at Lund University. Frank still has a keen interest in digital tools for research and keeps a blog on research policy and practices at www.researchaffairs.net