EcoSta 2019: Start Registration
View Submission - EcoSta2019
Title: Minimax optimality of sparse predictive density estimates Authors:  Gourab Mukherjee - University of Southern California (United States) [presenting]
Abstract: The problem of predictive density estimation is considered in a high-dimensional Gaussian model with sparsity constraints on the location parameters. The maximal risk of several spike-and-slab predictive density estimates is evaluated. Asymptotic minimaxity of these density estimates and the geometry of the least favorable prior distribution is discussed highlighting the contrasts with the minimax theory for point estimation of a multivariate normal mean under quadratic loss.