Graphical Models for Complex Health Data

This is a course at the intersection of statistics and machine learning, focusing on graphical models. In complex systems with many (perhaps hundreds or thousands) of variables, the formalism of graphical models can make representation more compact, inference more tractable, and intelligent data-driven decision-making more feasible. We will focus on representational schemes based on directed and undirected graphical models and discuss statistical inference, prediction, and structure learning. We will emphasize applications of graph-based methods in areas relevant to health: genetics, neuroscience, epidemiology, image analysis, clinical support systems, and more. We will draw connections in lecture between theory and these application areas. The final project will be entirely “hands on,” where students will apply techniques discussed in class to real data and write up the results.