A2048
Title: Sequential change detection with active sampling
Authors: Georgios Fellouris - University of Illinois, Urbana-Champaign (United States) [presenting]
Abstract: The problem of sequential change detection with active sampling is considered, where data are collected sequentially, an action is taken at each time instant, and an abrupt change occurs at an unknown time. Each action can be selected in real time based on the previously collected data and influences the distribution of the current observation both before and after the change. The objective is to jointly design an action policy and a stopping rule in order to detect the change as quickly as possible while controlling the false alarm rate. The pre-change distribution is completely specified, but a general composite post-change regime is considered. A scheme is presented that achieves first-order asymptotic optimality under Lordens minimax criterion as the mean time to false alarm and possibly the number of controls tend to infinity.
