The study of network dynamics, or how large numbers of individual components interact, is foundational to much of the research at SFI. Network theory has offered insight into everything from stock markets and ecosystems to disease spread and human social behavior. But there are still vexing challenges — problems that look simple but which prove to be quite hard — at the heart of network theory and models.
Incoming Program Postdoctoral Fellow George Cantwell, who is completing his PhD in physics at the University of Michigan, recently tackled one well-known flaw in network modeling that has persisted since the 1930s. A common tool, belief propagation, used for calculating network properties, only works on networks with no short loops. But most real-world networks have short loops.
That disconnect between models and their ability to rigorously fit to real-world data is the vexing challenge that drives Cantwell’s research. He would like to see theoretical research orient toward consequential real-world questions. The questions of greatest import for Cantwell, who also holds a BA in physics and philosophy from the University of Oxford, concern human sociology and psychology. “Many of the hardest and most interesting questions arise in this context,” he says. At SFI, Cantwell will be working with Professor Cris Moore on an NSF-funded project aimed at identifying when it is mathematically possible to pull patterns from large, noisy, high-dimensional real-world datasets.