Yanna Shen 
I am a Ph. D. student in the Intelligent Systems Program at the University of Pittsburgh, advised by Dr. Gregory Cooper. I am working as a Graduate Student Researcher at RODS Laboratory. My research interests include machine learning, graphical models, Bayesian statistics, and the investigation of these methods in addressing realworld problem. 
Suite M183 VALE, 200 Meyran Ave University of Pittsburgh Pittsburgh, PA 15260
Phone: (412)6486710 Email: shenyn@cs.pitt.edu 
Detecting anomalous events in data has important applications in domains such as disease outbreak detection, fraud detection, and intrusion detection. In a typical scenario, a monitoring system examines a sequence of data to determine if any recent activity can be considered as deviation from baseline behavior. Many anomalydetection algorithms, such as cumulative sum method, use frequentist statistical techniques. I am currently doing Bayesian modeling for anomaly detection. ►The goals of automated diseaseoutbreak detection systems are to detect disease outbreaks early, while exhibiting few false positives. ¡î I am now working on Bayesian modeling of unknown causes of events in the context of diseaseoutbreak detection. I developed a Bayesian hybrid detection algorithm that models and detects both known diseases (e.g., influenza and anthrax) by using informative prior probabilities and unknown diseases (e.g., a new, highly contagious respiratory virus that has never been seen before) by using relatively noninformative prior probabilities [3,7]. ¡î I developed a general Bayesian univariate anomalydetection algorithm that runs in linear time [9]. I intend to develop a multivariate version of this algorithm for monitoring multiple features of some event. ¡î I developed an efficient spatial Bayesian outbreakdetection algorithm that performs complete Bayesian model averaging over all possible spatial hypotheses, which we call SBMA [2]. I intend to develop a multivariate version of SBMA and apply SBMA to a wide variety of diseaseoutbreak scenarios. ►The objective when evaluating a diseaseoutbreak detection algorithm is to measure its accuracy (sensitivity and false alarm rate) and time to detection. However, little research has been done on estimating how well automated diseaseoutbreak detection systems augment traditional outbreak detection that is carried out by clinicians. It would be best to develop algorithms that are augmentative of, rather than redundant with, clinician detection. ¡î I developed a general mathematical framework for evaluating joint clinicianmachine detection of anomalies [1,4].
In the summer of 2008, I did research at Intel Research Pittsburgh through a summer internship there. I worked on Bayesian dynamic joint modeling of multiple perspectives in spatialtemporal inference problems [8]. 
1. Shen, Y., C. Adamou, J. N. Dowling, and G.F. Cooper, Estimating the joint disease outbreakdetection time when an automated biosurveillance system is augmenting traditional clinical case finding. Journal of Biomedical Informatics, 2008. 41(2): 224231. [PDF] 2. Shen, Y., W.K. Wong, J. Levander and G.F. Cooper, An outbreak detection algorithm that efficiently performs complete Bayesian model averaging over all possible spatial distributions of disease. Advances in Disease Surveillance 2007; 4:113. [PDF] 3. Shen, Y. and G.F. Cooper, A Bayesian biosurveillance method that models unknown outbreak diseases. In: Proceedings of Intelligence and Security Informatics: BioSurveillance 2007: 209215. [PDF] 4. Shen, Y., W.K. Wong, and G.F. Cooper, Estimating the expected warning time of outbreakdetection algorithms. Advances in Disease Surveillance 2006; 1:65. [PDF] 5. Lu, X., Q. Li, Z. Huang, Y. Shen and T. Yao, Towards ChineseEnglish sentence alignment based on statistical method. Journal of MINIMICRO Systems. 2004, Vol. 25, No. 6, 990992. 6. Zhang, L., X. Lu, Y. Shen and T. Yao, A statistical approach to extract Chinese chunk candidates from large corpora. In: Proceedings of International Conference on Computer Processing on Oriental Languages, 2003. Papers Submitted or in Preparation 7. Shen, Y. and G.F. Cooper, Bayesian modeling of unknown diseases for biosurveillance. Submitted to Artificial Intelligence in Medicine. 8. Denver Dash, Y. Shen, and Matthai Phillipose, AMP: Automating Multiviewpoint Perception. Submitted to NIPS¡¯08 Workshop: Learning from Multiple Sources. 9. Shen, Y. and G.F. Cooper, A Bayesian univariate anomalydetection algorithm. 10. Shen, Y. and G.F. Cooper, An efficient Bayesian model averaging algorithm for spatial diseaseoutbreak detection.

An outbreak detection algorithm that efficiently performs complete Bayesian model averaging over all possible spatial distributions of disease. 2007 International Society for Disease Surveillance Annual Conference, Indianapolis, Indiana USA, October 2007. [PPT] A Bayesian biosurveillance method that models unknown outbreak diseases. NSF Workshop on BioSurveillance Systems and Case Studies (BioSurveillance 2007), New Brunswick, New Jersey USA, May 2007. [PPT] Estimating the expected warning time of outbreakdetection algorithms. 2005 Syndromic Surveillance Conference, Seattle, Washington USA, September 2005. [PPT] 
2003 Fall 2004 Spring 2004 Fall 2005 Spring 2005 Fall 2006 Spring 2006 Fall 
Last update: Oct 2008
