"Cartography of Complex Systems"

Many complex systems need protection. Frustratingly, there is often no clear objective way to measure a system's health. The most common alternative— of waiting for parts of a system or outputs from it to degrade— can be too reactionary. This talk explores an easier alternative: comparing systems to each other. Similar to how machine learning methods can classify and predict without a mechanistic model, a set of inter-system distances can be used to create a map describing a system's behavior. These kinds of maps can then be used to predict risk and infer processes at play. I will explore an empirical example of this approach to understanding complex systems using water quality assessments of the Chesapeake Bay, and then a theoretical example of how these maps can be used to classify arbitrary network data according to their generating mechanisms.