David B. Dunson


Arts and Sciences Distinguished Professor of Statistical Science

My research focuses on the development of statistical and machine learning methodology motivated by complex and high dimensional data. Motivating applications include neuroscience, environmental health, epidemiology, inference from medical records, ecology, sports statistics, and recommender systems among others. A particular emphasis is on developing flexible Bayesian and probabilistic learning approaches for inferring and exploiting lower dimensional structure in high-dimensional and object/functional data. This includes novel latent factor models, autoencoders and geometric/manifold learning approaches. Important aspects include valid uncertainty quantification, interpretable inferences, scalable algorithm design, and adaptive approaches that increase the complexity of the specification as needed as size and complexity of the data increase.