Each semester, the Department of Statistics explores an advanced topic in statistical science in its Seminar on Statistical Theory (STAT-S 785). For the Fall 2025 semester, Professor Michael Trosset has organized a synchronous on-line course in which he and his long-time collaborator, Professor Carey Priebe in the Department of Applied Mathematics & Statistics at Johns Hopkins University, will present, discuss, debate, and speculate about a line of investigation that they have been pursuing for the past decade. Participants will include graduate students and faculty at IU and JHU, as well as North Carolina State University, University of Wisconsin, University of Maryland, University of Delaware, and University of Edinburgh. The seminar will offer graduate students a unique opportunity to observe and interact with a number of prominent researchers as they contemplate a unified set of topics at the frontier of modern statistics, machine learning, and artificial intelligence.
Professors Trosset and Priebe describe their research program as "manifold learning for subsequent inference". The fundamental objective of their work is developing practical methods and accompanying statistical theory for analyzing various high-dimensional data sets. Their methods attempt to learn enough about a data set's high-dimensional structure that its essential features can be represented and studied in a familiar low-dimensional space. Applications include (1) testing hypotheses about neuro-behavioral maps of Drosophila larvae, (2) statistical analyses of networks and time series of networks, and (3) studying the behavior of large language models in artificial intelligence.
