Abstract:
 Images (and videos) are very high dimensional signals that reside in a wide spectrum of heterogeneous subspaces of varying dimensions.   Traditionally, we have two mathematical tools for image modeling: (i) Markov random fields (Gibbs), originated from modern statistical physics for texture patterns in very high dimensional subspaces; and (ii) sparse coding and wavelet, originated from harmonic analysis for low dimensional subspaces. It is unknown in the literature how these two methods can be related to each other. In this talk, firstly, I will show two types of pure manifolds: (i) implicit manifold for high entropy patterns, like texture, modeled by MRF and (ii) explicit manifolds for low entropy patterns, like textons and image primitives, modeled by sparse coding. Secondly, I will present a unifying theory for learning probabilistic models by manifold pursuit through information projection, and show the two theories (MRF and sparse coding) work in two  image regimes (high and low entropy).  Thirdly I will discuss how these manifold are mixed to form middle entropy patterns and learn hybrid image templates for object categories. Finally I will discuss some open problems in image representation and machine learning: such as measuring the capacity and learnability of hypothesis space.

Biography：
Degrees
• 1996 Ph.D., Harvard University, Cambridge, MA
• 1994 M.S., Harvard University, Cambridge, MA
• 1991 B.S., University of Science and Technology of China, at Hefei, China
Appointments
• 2006 Professor, University of California at Los Angeles, Depts. of Statistics, Computer Science
• 2002 Associate Professor, University of California at Los Angeles, Depts. of Statistics, Computer Science
• 1998 Assistant Professor, Ohio State University, Depts. of Computer Science, Cognitive Science
• 1997 Lecturer, Stanford University, Dept. of Computer Science
• 1996 Post-doc, Brown University, Division of Applied Math.
Academic Honors
• 2008 J.K. Aggarwal Prize, the Int'l Association of Pattern Recognition. Citation
• 2007 Marr Prize honorary nomination, 11th ICCV at Rio, Brazil, for Object Modeling with Y. Wu etc.
• 2007 Changjiang Scholar, Ministry of Education, China.
• 2003 Marr Prize, 9th ICCV at Nice, France, for MCMC Inference for Image Parsing with Z. Tu etc.
• 2001 Young Investigator Award, Office of Navy Research.
• 2001 Sloan Fellow, Alfred P. Sloan Foundation.
• 2001 Career Award, National Science Foundation.
• 1999 Marr Prize honorary nomination, 7th ICCV at Corfu, Greece, for Texture Modeling with Y. Wu.
• 1995 Jury Prize, Harvard University.
• 1992 Harvard Fellowship, Harvard Graduate School of Art and Sciences.
• 1986-1991 Numerous undergraduate student awards in China
Editorial Service
• Editing Board, International Journal of Computer Vision (2004-- )
• Editing Board, IEEE Transactions on Pattern Analysis and Machine Intelligence (2005--2009 )
• Editing Board, Foundations and Trends in Computer Graphics and Vision (2004-- )