The aim of this project is to plant the seeds for an interdisciplinary institute devoted to foundations of data science at MIT. During the three years of Phase I, the project goal is to stimulate research and educational interactions between mathematics, statistics and theoretical computer science, both within MIT and in the research community at large. On the way, the team will also develop the organizational capabilities and visibility needed for the (potential) Phase II activities.
The project activities in Phase I will be organized around 5 semester-long themes. Each theme will be devoted to a topic at the intersection of at least two (and often three) TRIPODS areas, and will focus on catalyzing interactions between them. The specific themes are: statistical and computational tradeoffs, sub-linear algorithms and distribution testing, learning with complex structures, graphical models and exchangeability, and non-convex optimization. Within each theme, the PIs will organize numerous activities, including a workshop, a fall/spring school covering introductory material, and regular seminars. They will also support knowledge transfer activities, including office hours that will offer free help and consultation on algorithmic, mathematical and statistical aspects of data science. To facilitate the interactions between the participants, the project will support two project postdocs, who will play the role of connectors between different disciplines.