Personal Website of George V. Moustakides
Information
Scientific Activities
Publications
Teaching
Local Links
CS Rutgers
Activity
 Current research areas
       Sequential detection
       Machine Learning and Nonlinear Modeling
 Past research areas
 R & D projects
 Collaborations
 Invited talks
 Editorial & organizational activities
 
Current Research Areas
Sequential detection: Deciding sequentially between two hypotheses or detecting changes in the statistics of random processes are two problems with num­erous applications as intrusion detection, systems monitoring, quality control, image processing, biomedical engineering, com­mun­ica­tions, etc.
The two problems we are interested in are the following. In the first we are given sequentially observations and we are asked to decide between two or more different hypotheses (hypothesis testing problem). In the second problem our data initially have statistics according to some nominal model and at some unknown point in time these statistics change abruptly and we are asked to detect the change as soon as possible (change detection problem). The detection schemes we are interested in are sequential, in other words, at every time instant as new information arrives we perform a test to decide whether to stop and make a decision or continue receiving new data. Of course we are interested in schemes that do not exhibit frequent false alarms and detect the change asap. Depending on the a-priori knowledge we have for the statistics before and after the change, the difficulty of the problem varies.
Theoretical results concerning optimum schemes, even for the simple case where all statistics are completely known, are limited. The main difficulty of the problem stems from its sequential nature, that is, the need to apply a test every time new information arrives.
When the statistics are not exactly known then available techniques are basically ad-hoc and, at best, characterized by asymptotic optimality. As was previously mentioned obtaining optimum schemes for this problem is rather difficult. Asymptotically optimum schemes, on the other hand, seem to be more feasible and clearly of great practical impor­tance.
Our goal is to find optimum detection schemes for continuous and discrete time processes when the statistics are completely known. Our main emphasis is placed on the dependent observation data case and on the independent case but with multiple alternatives. We should mention that these problems have well known solutions in the fixed sample size case, but under the sequential setup they have been open for many years!
For more information please consult the relevant publications in  Sequential Techniques & Optimal Stopping 
   to top 
Machine learning and nonlinear modeling: We are interested in neural network models and in particular training methods that are not supervised. We target three categories of problems: In the first we are interested in data synthesis which is an area largely dominated by Generative Adversarial Networks (GANs). In the second we focus on obtaining probability densities based on neural network modeling for a given collection of data. And in the third we would like to extend the previous ideas to random signals and obtain Markov-type models, again based on neural network induced nonlinear transformations.
We would like to develop non-adversarial methods for the training of generative networks and use these ideas to construct neural-network-based statistical models for random data vectors and random signals.
For more information on adaptive estimation please consult the relevant publications in  Machine Learning and Nonlinear Modeling 
This is a completely new research area and we are looking into identifying problems that present theoretical and computational challenges.
to top
 
Past Research Areas
Event-Triggered Sampling
    Decentralized detection using event-triggered sampling
    State estimation using adaptive sampling
    Event based statistical signal processing
    For relevant publications please follow the link  Event-Triggered Sampling & Applications 
Physical Layer Communications
    Detection of access layer misbehavior in Wireless Networks
    Blind channel estimation in CDMA, OFDM and MC-CMDA
    BER estimation in optical communication systems
    For relevant publications please follow the link  Digital Communications & Networks 
Statistical Signal Processing
    Adaptive blind source separation
    Theoretical comparison methods based on stochastic approximation
    Convergence study of adaptive estimation algorithms
    Stabilization of fast estimation algorithms
    Low computational complexity estimation algorithms for echo cancellation and equalization
    For relevant publications please follow the link  Statistical Signal Processing 
Probabilistic and Statistical Methods in Databases
    Privacy preserving data mining
    Record Linkage
    For relevant publications please follow the link  Probabilistic & Statistical Methods in Databases 
1-D and 2-D Filter Design Techniques
    Design of two and multi-dimensional FIR filters using transformations
    Optimum one and two-dimensional FIR filters using L2 techniques
    FIR and IIR polynomial predictive filters
    For relevant publications please follow the link  Filter Design Techniques 
Off-line Detection of Changes
    Detection and diagnosis of changes in physical structures
    Local techniques for detecting changes
    For relevant publications please follow the link  Off-line Detection of Changes 
Detection and Estimation
    Robust detection of signals in dependent noise
    Robust filtering in dependent noise
    For relevant publications please follow the link  Detection & Estimation 
to top
 
Research & Development Projects
Quickest change detection techniques with signal processing applications (2015-2019, PI). Department of Computer Science, Rutgers University, USA. Financed by the National Science Foundation, USA. Rutgers Budget: 467,186$ (Total Budget: 1,121,161$).
Event-triggered sampling: Application to decentralized detection and estimation (2011-2015, co-PI). Department of Electrical Engineering, Columbia University, USA. Financed by the National Science Foundation, USA. Budget: 952,747$.
Adaptive signal processing algorithms for digital hearing aids (2008-Today, PI). Department of Electrical and Computer Engineering, University of Patras, Greece. Financed by SoundWorld Solutions, USA. Budget: 15,000€/year.
Robust rapid change-point detection in multi-sensor data fusion and behavior research (2008-2009, co-PI). Georgia Institute of Technology, School of Industrial and Systems Engineering, Atlanta, GA, USA. Financed by the Air Force Office of Scientific Research (AFOSR). Budget: 200,000$.
Estimating performance in optical communication systems (2002-2003, PI). Institut National de Recherche en Informatique et en Automatique (INRIA), Rennes, France. Financed by ALCATEL. Budget: 50,000FF.
Adaptive techniques for wireless multiple-access communications (1998-1999). Department of Computer Engineering and Informatics, University of Patras, Greece. In collaboration with the Electrical Engineering departments of Princeton University and University of Pennsylvania, USA. NATO Scientific Exchange Programme: Collaborative Research Grant. Budget: 12,000$.
Development and study of adaptive estimation algorithms for telecommunication applications (1995-1997, PI). Department of Computer Engineering and Informatics, University of Patras, Greece. Financed by the Greek General Secretariat for Research and Technology. Budget: 100,000Drh.
Automatic finite element model generation of objects using image processing techniques (1992-1995, PI). Computer Technology Institute, Patras, Greece. In collaboration with Agricultural University of Athens, Greece, University of Leuven, Belgium, University of Swansea, Great Britain and University of Ilmenau, Germany. Financed by the European Union through the Human Capital and Mobility Programme. Budget: 100,000ECU.
Detection of visually evoked cortical potentials (1991-1992, PI). Department of Computer Engineering and Informatics, University of Patras, Greece. Financed by the Greek General Secretariat for Research and Technology. Bilateral collaboration program with the department of Biomedical Engineering, University of Ilmenau, Germany. Budget: 20,000Drh.
Development of a PC based visual evoked potentials station for noninvasive medical diagnosis (1991-1992, PI). Department of Computer Engineering and Informatics, University of Patras, Greece. Financed by the Greek General Secretariat for Research and Technology. Budget: 20,000Drh.
Development of a DSP based bio-impedance monitoring station (1989-1990, PI). Computer Technology Institute, Patras, Greece. Financed by the Greek General Secretariat for Research and Technology. Bilateral collaboration program with the department of Biomedical Engineering, University of Ilmenau, Germany. Budget: 20,000Drh.
Fast estimation algorithms for echo cancellation in communication channels (1989-1990). Department of Computer Engineering and Informatics, University of Patras, Greece. Financed by the Greek General Secretariat for Research and Technology. Budget: 20,000Drh.
Signal processing techniques for detecting and localizing fatigues in off-shore platforms (1983-1986). Institut National de Research en Informatique et en Automatique (INRIA), Rennes, France. Financed by Elf-Aquitaine. Budget: 200,000FF.
to top
 
Collaborations
University of Patras : (2018-Today), Computer Engineering and Informatics Department, Patras, Greece.
(Collaboration with Prof. E. Psarakis on Non-adversarial generative networks).
Rutgers University: (2017-Today), Psychology department, Piscataway, NJ, USA.
(Collaboration with Prof. E. Torres on Early detection of events in patients with epilepsy).
Georgia Institute of Technology: (2017-Today), School of Industrial and Systems Engineering, Atlanta, Georgia, USA.
(Collaboration with Prof. Y. Xie on Subspace change detection).
Georgia Institute of Technology: (2016-2018), School of Industrial and Systems Engineering, Atlanta, Georgia, USA.
(Collaboration with Prof. Y. Mei on Sequential estimation).
University of Illinois: (2014-Today), Coordinated Science Lab, Department of Electrical and Computer Engineering, Urbana-Champaign, IL, USA.
(Collaboration with Prof. V. Veeravalli on Sequential change detection).
Rutgers University: (2013-Today), Department of Computer Science, CBIM Lab, Piscataway, NJ, USA.
(Collaboration with Prof. D. Metaxas on Anomaly detection and Sparse representation).
University of Illinois: (2013-Today), Statistics Department, Urbana-Champaign, IL, USA.
(Collaboration with Prof. G. Fellouris on Sequential change detection).
University of Southern California (USC): (2010-2012), Department of Mathematics, Los Angeles, CA, USA.
(Collaboration with Dr. G. Fellouris on Decentralized sequential detection).
Columbia University: (2009-2013), Department of Electrical Engineering, New York, NY, USA.
(Collaboration with Prof. X. Wang on Application of event-triggered sampling to DSP and Stochastic Control).
SoundWorld Solutions: (2008-2014), Park Ridge, IL, USA.  website   youtube (CBS News) 
(Collaboration with Dr. S. Basseas on Signal processing for Hearing aids).
University of Southern California (USC): (2008-2014), Department of Mathematics, Los Angeles, CA, USA.
(Collaboration with Dr. A. Tartakovsky on Sequential changepoint detection).
Georgia Institute of Technology: (2008-2010), School of Industrial and Systems Engineering, Atlanta, Georgia, USA.
(Collaboration with Prof. Y. Mei on Sequential changepoint detection).
Columbia University: (2007-2010), Department of Statistics, New York, NY, USA.
(Collaboration with Grad. Stud. G. Fellouris on Decentralized sequential detection).
University of Maryland: (2005-2006), Institute for System Research, College Park, MA, USA.
(Collaboration with Prof. J. Baras and Dr. M. Rabi on Event-triggered sampling).
University of Thessaly: (2004-2010), Department of Electrical & Computer Engineering, Volos, Greece.
(Collaboration with Prof. V. Verykios on Statistical techniques for databases).
Princeton University: (1997-1999), Department of Electrical Engineering, Princeton, NJ, USA.
(Collaboration with Prof. H.V. Poor on Multiuser communications).
Ilmenau Technical University: (1990-1995), Institute of Biomedical Engineering and Informatics, Ilmenau, Thuringen, Germany. (Collaboration with Prof. G. Henning on Visual evoked potential).
to top
 
(conference session invitations are not included) Invited Talks (conference session invitations are not included)
Sequential change detection, Plenary talk, IEEE International Symposium on Information Theory (ISIT), Paris, France, July 2019.  website 
Detecting changes in Markov processes, Plenary talk, 6th International Workshop on Sequential Methodologies (IWSM), University of Rouen, Rouen, France, June 2017.  slides   pdf 
Metrics and optimum tests in sequential change-detection, Industrial and Systems Engineering department, Georgia Institute of Technology (GaTech), Atlanta, USA, December 2016.  slides    pdf 
Sequential detection of changes: Metrics and optimum tests, Statistics department, Wharton School, University of Pennsylvania, Philadelphia, USA, October 2016.  slides    pdf 
Sequential detection of a first-entry-to-a-set, France Research Center, Huawei Technologies, Paris, France, April 2016.  slides 
Adaptive algorithms, Digital Technology Center, Electrical and Computer Engineering, University of Minnesota, Minneapolis, MN, USA, January 2015. (Two one-hour lectures).
Sequential detection of changes: An Overview, Digital Technology Center, Electrical and Computer Engineering, University of Minnesota, Minneapolis, MN, USA, January 2015.  slides 
Sequential detection of changes: Overview and recent results, Electrical Engineering, Princeton University, Princeton, NJ, USA, January 2014.  slides 
Sequential change detection: Brief overview, Command, Control and Interoperability Center for Advanced Data Analysis, Rutgers University, New Brunswick, NJ, USA, December 2013.  slides 
Sequential detection and system identification, Statistics Department, University of Connecticut, Storrs, CT, USA, November 2013.  slides 
Optimal routing of autonomous vehicles in stochastic environments, Computational Biomedicine, Imaging and Modeling Center, Computer Science Department, Rutgers University, NJ, USA, December 2012.  slides 
Sequential detection: Overview and open problems, Advanced Network Colloquium, Institute for Scientific Research (ISR), University of Maryland, College Park, MA, USA, December 2011.  slides    pdf    video 
Optimum joint detection and estimation, application to MIMO radar, Department of Electrical Engineering, University of Southern California (USC), Los Angeles, CA, USA, January 2011.  slides    pdf 
Sequential rate change detection in Poisson processes, Department of Statistics and Applied Probability, University of California, Santa Barbara, USA, January 2011.  slides    pdf 
Joint detection and estimation, application to MIMO radar, Department of Electrical Engineering, Bilkent University, Ankara, Turkey, September 2010.  slides 
Optimum GLR tests, Probability & Statistics seminar, Department of Mathematics, University of Southern California (USC), Los Angeles, CA, USA, January 2009.  slides    pdf 
Finite sample size optimality of GLR tests, Electrical Engineering department, Columbia University, New York, NY, USA, September 2008.  slides    pdf 
Finite sample size optimality of GLR tests, Statistics seminar, School of Industrial and Systems Engineering, Georgia Institute of Technology (GA Tech), Atlanta, GA, USA, August 2008.  slides    pdf 
Asynchronous random sampling for decentralized detection, Applied Math seminar, Department of Mathematics, University of Southern California (USC), Los Angeles, CA, USA, February 2008.  slides    pdf 
Change-time models and performance criteria for the problem of sequential change detection, Probability & Statistics seminar, Department of Mathematics, University of Southern California (USC), Los Angeles, CA, USA, February 2008.  slides    pdf 
Decentralized sequential hypothesis testing and change detection, Laboratoire de Modélisation et de Sûreté de Systèmes, Université de Technologie de Troyes, Troyes, France, January 2008.  slides 
Sequential techniques for hypothesis testing and change detection, School of Electrical Engineering, The Royal Institute of Technology (KTH), Stockholm, Sweden, October 2007.  slides    pdf 
Sequential detection of changes (mini summer-course of nine lectures), Department of Statistics, Columbia University, NYC, NY, USA, May-June 2007.  slides  of the first lecture.   pdf 
The Poisson disorder problem, Department of Operations Research and Financial Engineering, Princeton University, Princeton, NJ, USA, February 2006.  slides    pdf 
Performance evaluation of CUSUM tests, Department of Statistics, Columbia University, New York, USA, March 2003.  slides 
Optimum CUSUM tests for detecting changes in continuous time processes, 11-th Annual Applied Probability Day, The Center for Applied Probability, Columbia University, New York, USA, March 2003. (Six speakers are invited to this annual event).  slides    pdf 
Optimum sequential procedures for detecting changes in processes, Université du Maine, Faculté des Sciences, Laboratoire de Statistique et Processus, Le Mans, France, December 2002.  slides    pdf 
Adaptive algorithms for blind separation of dependent sources, Department of Mathematics, Université Joseph Fourier - Grenoble I, Grenoble, France, January 2002.  slides    pdf 
Optimum adaptive algorithms for blind source separation, Institut d'informatique, Université de Neuchâtel, Neuchâtel, Switzerland, April 2001.  slides 
Decision directed algorithms for multiuser detection, Institut de Recherche en Informatique et en Automatique (INRIA - IRISA), Rennes, France, November 2000.  slides 
Adaptive algorithms for blind source separation, Signal Processing Lab., Department of Electrical Engineering, University of Pennsylvania (UPENN), Philadelphia, USA, July 2000.  slides 
Optimum adaptive signal processing algorithms, Laboratoire de Modélisation et de Sûreté de Systèmes, Université de Technologie de Troyes, Troyes, France, June 1999.  slides 
New developments in sequential analysis, Department of Statistics, Hebrew University of Jerusalem, Jerusalem, Israel, March 1998.  slides 
to top
 
Editorial & Organizational Activities
IEEE Transactions on Information Theory: Inaugural Associate Editor for Sequential Methods, 2016-2020.
International Workshop on Applied Probability, IWAP 2018, 18-21 June, Budapest, Hungary: Co-organizer with V. Veeravalli of the invited session "Sequential Methods".
Conference on Information Sciences and Systems, CISS 2018, 21-23 March, Princeton, USA: Co-organizer with V. Veeravalli of the invited session "Sequential Methods for Signal Processing and Control".
2016 Asilomar Conference on Signals, Systems and Computers, 6-9 November 2016, Pacific Grove, CA, USA: Co-organizer with V. Veeravalli of the session "Sequential Signal Processing".
International Workshop in Sequential Methodologies, IWSM 2015, 22-25 June 2015, Columbia University, NYC, USA: Co-organizer with A. Tartakovsky of the session "Tribute to Alexander Novikov: Sequential Tests and Estimation". Co-organizer with V. Veeravalli of the session "Sequential/Quickest Change Detection: Theory and Applications".
IEEE Transactions on Information Theory: Associate Editor for Detection and Estimation, 2011-2014.  pdf 
International Workshop on Sequential Methods and their Applications, 4-8 June 2012, University of Rouen, France: Scientific Committee.  pdf 
International Workshop in Sequential Methodologies, IWSM 2011, 14-16 June 2011, Stanford University, Stanford, USA: Organizer of the session "Applications of Sequential Detection and Optimal Stopping".  pdf 
International Workshop on Applied Probability, IWAP 2010, 5-8 July 2010, Universidad Carlos III of Madrid, Colme­narejo Campus, Madrid, Spain: Co-organizer with I. Nikiforov and A. Tartakovsky of the invited session "Sequential Detection & Estimation I,II,III,IV".  pdf 
International Workshop on Sequential Methodologies, IWSM 2009, 15-17 June 2009, Université de Technologie de Troyes, Troyes, France: Steering committee, Program committee (Chair), Local organization committee, Co-organizer with A. Tartakovsky of the invited session "Sequential Testing and Optimal Stopping I,II,III" and with I. Nikiforov of the invited session " Sequential change-point detection and isolation".  pdf 
International Workshop in Sequential Methodologies, IWSM 2007, 22-25 July 2007, Auburn University, Auburn, Ala­bama, USA: Steering committee.
EURASIP Journal on Applied Signal Processing, vol. 2007, K. Berberidis, B. Champagne, G.V. Moustakides, H.V. Poor, P. Stoica, Special issue on Advances in subspace-based techniques for Signal Processing and Communi­ca­tions: Guest editor.  editorial 
European Signal Processing Conference, EUSIPCO 1998, 8-11 September 1998, Rhode, Greece: Organizing commit­tee (Tutorials).  pdf 
to top