Princeton University

School of Engineering & Applied Science

Sun-Yuan Kung

Professor of Electrical Engineering


Room: B230 Engineering Quadrangle
Phone: 609-258-3780
Email: kung@princeton.edu

Education

  • Ph.D., Stanford University, 1977
  • M.S., Electrical Engineering, University of Rochester, 1974
  • B.S., Electrical Engineering, National Taiwan University, 1971

My research group has been focused on development of  numerical machine learning method based on optimization formulation over nonlinearly expanded vector spaces.  By automatic machine learning, we can find dimension-reduced feature subspace such that the subspace  would be least lossy for the intended utility user but most lossy for the potential privacy intruder. This leads to a novel ``compressive privacy” approach to delivering the intended  utility (such as pattern recognition or classification) while preserving personal privacy (such as de-identification).  Quantitatively, the relevance of a nonlinear space can be calibrated by a DI-metric where DI stands for  ``discriminant information" or ``differential information", which proves to  be instrumental  for (1) the development of a ``Discriminant Component Analysis” (DCA) for optimal subspace representation and (2) the selection of optimal kernel functions in  multi-kernel machine learning models.   In terms of novel machine learning models, we have developed  component analysis methods (DCA and KDCA) and multi-kernel classifiers.  In addition, we are exploring a  hybrid kernel-in-deep learning network (KiDNet)  to seamlessly embed kernel features into deep learning neural networks. In terms of applications,  our models have already been successful applied to several demonstrations for  privacy-preserving data mining systems per our DARPA Brandeis Program, including DIAC, DSAC, DIAR,  and FOID.
 
 

Honors and Awards

  • Life Fellow of IEEE, for contribution to VLSI signal processing and neural networks, 2016
  • The Third Millennium Medal, IEEE, 2000
  • Best Paper Award, IEEE Signal Processing Society (1996)
  • Distinguished Lecturer, IEEE Signal Processing Society, (1994)
  • Honorary Professorship, Central China Science & Technology University, (1994)
  • Technical Achievement Award, IEEE Signal Processing Society, (1992)
  • Fellow of IEEE, for contribution to VLSI signal processing and neural networks, (1988)
  • Sino-US Exchange Scientist, National Academy of Science, 1987

Selected Publications

  1. S. Y. Kung, ``Compressive Privacy: From Information/Estimation Theory to Machine Learning”, Invited Lecture-Note, pp. 94-112,  IEEE Signal Processing Magazine, Jan.  2017.

  2. S.Y. Kung, "Discriminant component analysis for privacy protection and visualization of big data",  J. of Multimedia Tools and Application, 2017.

  3. S. Y. Kung, T. Chanyaswad, J. Morris Chang, and P. Wu, “Collaborative PCA/DCA learning methods for compressive privacy,” ACM Trans. Embed. Computer Systerms. (Special Issue on Effective Divide-and-Conquer, Incremental, or Distributed Mechanisms of Embedded Designs for Extremely Big Data in Large-Scale Devices),   2017.

  4. S..Y .Kung, ``A Compressive Privacy approach to Generalized Information Bottleneck and Privacy Funnel problems", Journal of the Franklin Institute, DOI: 0.1016/j.jfranklin.2017.07.002, online July2017, Elsiever.

  5. Yinan Yu, K.  Diamantaras, T. McKelvey, S. Y. Kung, "CLAss-specific Subspace Kernel Representations and Adaptive Margin Slack Minimization for Large Scale Classification",  the IEEE Transactions on Neural Networks and Learning Systems (TNNLS-2016-P-6571), in press.

  6. S. Y. Kung, “Machine Learning for Data Representation, Mining, and Privacy”,

    Springer/Nature,  in preparation.

  7. S.Y.Kung, "Kernel Methods and Machine Learning", Cambridge University Press,  615 pages,  2014.

  8. S.Y. Kung, M.W. Mak and S.H. Lin, "Biometric Authentication – A Machine Learning Approach", Prentice-Hall  Information and System Science Series, (T. Kailath, Series Editor) , 476 pages, 2005.

  9. K.I. Diamantaras and  S.Y. Kung,  ``Principal Component Neural Networks: Theory and Applications'', Adaptive and Learning System Series, John Wiley & Sons, Inc., New York, 476 pages, 1996.  

  10. S.Y. Kung, “Digital Neural Networks”,   Prentice-Hall  Information and System Science Series, (T. Kailath,  Series Editor),  440 pages, 1993.

  11. S.Y. Kung,  "VLSI Array Processors,"  Prentice-Hall  Information and System Science Series, (T. Kailath,   Series Editor), 667 pages, 1987.