*********************************
There is now a CONTENT FREEZE for Mercury while we switch to a new platform. It began on Friday, March 10 at 6pm and will end on Wednesday, March 15 at noon. No new content can be created during this time, but all material in the system as of the beginning of the freeze will be migrated to the new platform, including users and groups. Functionally the new site is identical to the old one. webteam@gatech.edu
*********************************
TITLE: Mixture procedure for sequential multi-sensor changepoint detection
SPEAKER: Dr. Yao Xie, Duke University
ABSTRACT:
Multi-sensor change-point detection, where sensors are distributed to monitor an abrupt emergence of a signal, has attracted much interests recently due to its potential applications in sensor networks, distributed anomaly detection, epidemiology, etc. The emergence of the change-point may abruptly change the distribution of the sensor observations. The goal is to detect such a signal as soon as possible after it occurs, and rarely make false alarm when there is not one. A challenge is that the subset of sensors affected by the change-point is typically unknown and small, i.e., the fraction p of sensors affected is very small. Without taking this signal sparsity into account, we may end up including much noise from the unaffected sensors and hurt detection performance.
We model this sparsity by assuming that each sensor has a probability p0 to be affected by the changepoint, and p0 is a guess for p. Based on this model, we develop a mixture procedure for monitoring parallel streams of data for a change-point that affects only a subset of them, without assuming a spatial structure relating the data streams to one another. Observations are assumed initially to be independent standard normal random variables. After a change-point the observations in a subset of the streams of data have unknown non-zero mean values. The procedure we study uses specific generalized likelihood ratio statistics. An analytic expression is obtained for the average run length (ARL) when there is no change and is shown by simulations to be very accurate. Similarly, an approximation for the expected detection delay after a change-point is also obtained. Numerical examples are given to compare the suggested procedure to other procedures for unstructured problems and in one case where the problem is assumed to have a well defined geometric structure. We discuss sensitivity of the procedure to the assumed value of p0, and extend this to a parallel mixture procedure. Finally, we apply the mixture procedure for real time solar flare detection in videos from the Solar Data Observatory.
Bio:
Dr. Yao Xie received the B.S. degree in Electrical Engineering from the University of Science and Technology of China (USTC), Hefei, China, in 2004, the M.S. degree in Electrical Engineering from the University of Florida, Gainesville, FL, in 2006, and the Ph.D. degree in Electrical Engineering (minor in Mathematics) from Stanford University, Stanford, CA, in 2011. She is currently a postdoctoral Research Scientist with the Electrical and Computer Engineering Department, Duke University, Durham, NC. Her current research interests include statistical sequential methods for signal and information processing, compressed sensing, optimization, and their applications in wireless communications, sensor networks, and medical imaging.
