Mauro Maggioni is interested in mathematical techniques for analyzing, modeling, and extracting information from large data sets that lead to smarter machine learning algorithms and scientific discoveries. This work creates the foundation to build generalizable, interpretable predictive models and thus has implications for a broad range of fields.

Maggioni’s research is focused on the analysis of high-dimensional data, graphs, and networks. Specifically, he analyzes and exploits the geometry of large data sets in order to train machines to predict patterns in this data. These hidden geometric structures are pervasive, appearing in a broad range of data types, from images to text documents to trajectories of complex dynamical systems. Maggioni applies a range of techniques to the study of physical systems, for instance for anomaly detection or to learn optimal policies for automated agents. He has recently applied this expertise to improving computational detection of chemicals in hyperspectral images and to training computers to recognize objects in art drawings.

Maggioni joined Johns Hopkins University as a Bloomberg Distinguished Professor in 2016 from Duke University.