Hierarchical Discriminant Regression (HDR) for Multimodal
                    Development and Integration 

                          Juyang Weng
                      Associate Professor
          Department of Computer Science and Engineering
         Michigan State University, East Lansing, MI 48824
                       weng@cse.msu.edu 
                 http://www.cps.msu.edu/~weng/

How could a machine achieve visual invariance for position, size and
orientation?  How could it achieve auditory invariance for speaker
independence and time warping?  How could it effectively integrate
different sensing modalities?  In this talk, we will examine some
studies in neuroscience and psychology which give thought provoking
evidence.  Next, a hierarchical discriminant regression (HDR) is
presented as a general purpose classifier and regressor, for
incremental autonomous generation of architecture and representation
for memory self-organization and updating.  It is compared with neural
networks, supporting vector machine (SVM), and other tree classifiers
such as CART and C5.0 for high dimensional pattern recognition in face
recognition, hand written numeral recognition and other pattern
recognition problems.  Further, HDR is used for developing visual and
auditory capabilities by SAIL robot through real-time online
interactions with the real physical world.  Three modes of learning
are integrated by the SAIL developmental program, supervised learning,
reinforcement learning and the new communicative learning.  In the new
communicative learning, language acquisition (language learning) and
teaching using language are carried out in the same autonomous mode.
The talk will show a video tape that demonstrates how the SAIL robot
learns simple spoken instructions and how humans use spoken commands
to teach the robot to manipulate objects and perform vision-guided
navigation.

--------------------- short bio ------------------------------
About the Speaker:

Juyang Weng received the BS degree from Fudan University, Shanghai,
China in 1982, and the M.S. and Ph.D. degrees from University of
Illinois, Urbana-Champaign, USA, in 1985 and 1989, respectively, all
in computer science.  Currently, he is an associate professor at
Department of Computer Science and Engineering, Michigan State
University, East Lansing, Michigan USA.  He is an originator of a new
direction called autonomous development approach to artificial
intelligence.  Currently, he serves as an associate editor of IEEE
Transactions on Pattern Analysis and Machine Intelligence.  He was a
program co-chair for NSF/DARPA Workshop on Development and Learning
held April 5-7, 2000, at Michigan State University
(http://www.cse.msu.edu/dl/) and is a program co-chair for 2nd
International Conference on Development and Learning (ICDL'02) to be
held at MIT, June 12-15, 2002 (http://www.egr.msu.edu/icdl02/).  His
current research interests include computer vision; learning robots
for vision, speech, language, manipulation, and navigation;
human-machine interface, visual reality and artificial intelligence.