B0452
Title: Sufficient dimension reduction with simultaneous region selection for high dimensional tensors
Authors: Shanshan Ding - University of Delaware (United States) [presenting]
Abstract: A unified framework is introduced for sufficient dimension reduction (SDR) on high-dimensional and tensor-valued data. SDR is known to be a powerful tool for achieving data reduction and visualization in statistical and machine-learning problems. Robust nonparametric SDR methods are proposed for data with high-dimensional tensor-valued features under weak assumptions, and develop a new framework for high-dimensional tensor SDR problems with theoretical guarantees. Promising applications are demonstrated through simulations and real data analysis on neuroimaging data.