Title: Efficient Active Quickest Detection for Streaming Data Under Sampling Control

Date: Monday, June 17th

Time: 10:00 AM – 12:00 PM ET

Location: Groseclose 403

Meeting Link: https://gatech.zoom.us/j/92807046631

Name: Qunzhi Xu

Industrial Engineering PhD Candidate

H. Milton Stewart School of Industrial and Systems Engineering
Georgia Institute of Technology

Committee

Dr. Yajun Mei  (Advisor), School of Industrial and Systems Engineering, Georgia Institute of Technology

Dr. Jie Chen, Department of Population Health Sciences, Augusta University

Dr. Roshan V. Joseph , School of Industrial and Systems Engineering, Georgia Institute of Technology

Dr. George V. Moustakides, Department of Electrical and Computer Engineering, University of Patras

Dr. Jianjun Shi, School of Industrial and Systems Engineering, Georgia Institute of Technology

Dr. Tuo Zhao, School of Industrial and Systems Engineering, Georgia Institute of Technology

Abstract

Quickest detection has a wide range of real-world applications in industrial quality control, biosurveillance, network security, etc. Under a general setting, there are p local streams in a system, and at some unknown time ν, an occurring event impacts s of the available streams by changing the distribution of their samples. In many applications, one often faces the sampling control constraint in the sense of allowing only to sample from q of the p local streams at each time instant. We call this “Active Quickest Detection”. The objective of active quickest detection is to decide how to adaptively sample partial data from these p local streams and how to use the observed partial data to raise a global alarm as quickly as possible once the change occurs subject to both the false alarm and sampling control constraints. This dissertation focuses on making comprehensive progress on methodology, theory, and application of active quickest detection problem to multi-stream data under the sampling or resource constraints. Our specific research aims are to design new algorithms with theoretical guarantees and develop an asymptotic optimality theory to characterize sharp information bound.