Finding Patterns in Large Networks

Many systems of scientific interest can be represented as networks, including the world wide web and the Internet, genetic and metabolic networks, social networks both on- and off-line, and many others. With the growing availability of large-scale data sets describing some of these networks, we face the challenge of extracting scientific sense from their often complex patterns of connections. This talk will present two new approaches to the analysis of network data, one focusing on spectral methods and the other on techniques from machine learning and statistics. Both allow us to take large networks and find regularities among their connections, shedding light on how they might have grown or been designed, and on the relation between form and function in networked systems.