Adapting
Computational Intelligence to Large Data Sets
Professor Larry Hall
Department of Computer Science and Engineering
ENB 118 University of South Florida 4202 E. Fowler Ave. Tampa, Fl 33620-9951
[hall@csee.usf.edu] Ph. (813) 974-4195 FAX: (813) 974-5456
Abstract
Computational intelligence approaches can arbitrarily closely approximate
functions, e.g. neural networks and fuzzy systems. They can do effective
searches of a global space for optimization e.g. evolutionary computation.
Today very large sets of data are being collected. Some of them have class
labels, for example instances of cellular telephone fraud. Others, do not
have labels, for example pixels in a scene contained in an image. When data
sets grow large, they pose a challenge to computational intelligence
techniques to build models in a timely fashion. This talk will discuss
distributed approaches to dealing with very large data sets. It will also
discuss how to deal with data sets in which the class of interest is quite
small (examples of fraud, for instance). Methods of applying neural network
models to data sets with many examples and potentially skewed class
distributions will be discussed. Also, fuzzy clustering for very large data
sets will be discussed. Some examples will be given from the domain of
protein secondary structure prediction among other data sets. Arguments for
distributed learning which results in ensembles of classifiers rather than
subsampling large data sets will be presented.
Speaker Biography
Prof. Hall's research interests lie in hybrid
computational intelligence systems, machine learning, pattern recognition
and integrating AI into medical image processing. The exploitation of
imprecision with the use of fuzzy logic in pattern recognition, AI and
learning is a research theme. Distributed learning systems are another area
of interest. He has written over 50 journal papers and a number of
conference papers. He co-edited the 1994 joint North American Fuzzy
Information Processing Society (NAFIPS), IFIS and NASA conference
proceedings and the 1998 proceedings, and IFSA/NAFIPS'03 conference
proceedings. He is a fellow of the IEEE, a past president of NAFIPS. Also,
associate editor for IEEE Transactions on Fuzzy Systems, and the
International Journal of Intelligent Data Analysis. Fuzzy Logic. He is the
Editor-in-Chief for the IEEE Transactions on Systems, Man and Cybernetics,
Part B.
Perspectives on Intelligent Control of robotic system and Mechatronic
Systems
Professor
Toshio Fukuda
President,
IEEE Nanotechnology Council
Department of
Microsystem Engineering, Nagoya University Furo-cho, Chikusa-ku
Nagoya
464-8603, JAPAN. Phone: +81-52-789-4478, FAX: +81-52-789-3115
E-mail: fukuda@mein.nagoya-u.ac.jp
[http://www.mein.nagoya-u.ac.jp],
[http://www.huenet.org/],
[http://www.ieee-nano.org]
[http://www.mein.nagoya-u.ac.jp/MHS],
[http://www.mein.nagoya-u.ac.jp/maze]
Humanoid Robot HanSaRam: Schemes for ZMP compensation
Abstract:
The
purpose of this talk is to give an overview of recent progress and
development in humanoid robot, HanSaRam series. HanSaRam is a humanoid robot
undergoing continual design and development in the Robot Intelligence
Technology (RIT) Laboratory at KAIST. This talk also presents the
experimental results of the ZMP compensation in the walking and standing
posture of HSR-IV. During walking motion, proportional ZMP compensation
scheme is employed. A novel two mode Q-learning, which extends
standard Q-learning, is used to compensate ZMP in the standing posture. In
the proposed two mode Q-learning, both the success and failure experiences
of an agent is used for fast convergence. The effectiveness of two mode
Q-learning is verified through real experiment. Also, the recently developed
humanoid robot, HSR-V is introduced in this talk to give an idea about the
future directions of our ongoing research.
Professor Kim,
Jong-Hwan
President: FIRA
[
