Title:
Quadratic Neural Unit and Quadratic Neural Network: A Compromise for Industrial Applications
Date:
Monday, June 21, 2010 at 2:30 PM – 3:30 PM
Location:
E2 – 393 EITC, Fort Garry Campus, University of Manitoba
Speaker:
Prof. Ivo Bukovsky, PhD
Department of Instrumentation and Control Engineering
Czech Technical University in Prague
Prague, Czech Republic
Abstract:
This talk will first summarize issues related to an adaptive methodology for evaluation of complicated time series by low-dimensional neural architectures. Then, it will introduce a Quadratic Neural Unit (QNU) as a promising low-dimensional neural architecture. The quadratic nature of the QNU assures sufficiently high quality of nonlinear approximation, while the QNU’s convexity facilitates a simple optimization process in the vicinity of a global minimum. Next, a Quadratic Neural Network (QNN) with sequential learning will be introduced to cope with a large number of neural weights in the QNU, corresponding to a large number of input samples. This non-conventional neural architecture benefits from the convex (quadratic) attributes, while maintaining its ability to converge towards the global minimum. Finally, experimental results from the QNU and QNN in control and other applications will be discussed and compared to conventional neural networks.
Speaker Bio:
Dr. Ivo Bukovsky’s major research interests are the development of new neural architectures, adaptive methodologies, and their applications to real dynamical systems. He graduated from Czech Technical University in Prague where he received his Ph.D. in the field of Control and System Engineering in 2007. During his PhD studies, Ivo spent seven months working as a visiting researcher under supervision of Dr. M.M. Gupta at the Intelligent Systems Research Laboratory at the University of Saskatchewan within the frame of the NATO Science Fellowships Award. His Ph.D. thesis on nonconventional neural units and adaptive approach to evaluation of complex systems was recognized by the Verner von Siemens Excellence Award 2007. For six months in 2009, Ivo worked on research and biomedical applications under the Matsumae International Foundation at the Cyberscience Center at the Tohoku University in Japan. Ivo has become IEEE CIS Neural Network Technical committee in 2007 and IEEE CIS Student Activity Subcommittee in 2010.
Cost:
Free, All are welcome.
Contact:
For questions or more information contact Witold Kinsner at 474-6490.
