Keynote Speaker Professor de Figueiredo to Discuss
Intelligent Multi-Media Communication
December 7, 2006 – Attracting attendees and researchers from
across the world, the Institute of Electrical and Electronics Engineers
with the University of California, Irvine, hosted the IEEE International Symposium on Multimedia (ISM2005) Dec.
12–14 in Irvine, CA. The symposium featured invited keynote
speakers, including UC Irvine’s Rui J.P. de Figueiredo, professor
emeritus in Electrical
Engineering and Computer Science, who discussed emerging trends and
regarding intelligent multi-media communication.
This ISM2005 symposium is an international forum created to
exchange information regarding advances in the state-of-the-art and practice of
multimedia computing, as well as identify the emerging research topics and
define the future of multimedia computing.
The technical program at ISM2005 consisted of invited
talks, paper presentations, and panel discussions. The event was hosted at the Hyatt Regency
Irvine, and all UC Irvine colleagues and faculty were invited to attend.
The event website is http://ism2005.eecs.uci.edu/index.html.
PROFESSOR DE FIGUEIREDO’S KEYNOTE ABSTRACT:
Intelligent multi-media communication (IMMC) will enable a
machine to communicate with a human or with another machine in the same way you
and I do. For this purpose, a powerful abstraction is needed to model, design,
and implement machines that are capable of detecting, classifying, and
interpreting complex events present in the multi-media signal.
With this motivation, we will present and explain, in simple
terms, a rigorous abstract approach to IMMC in human/machine systems, whereby
the machine is modeled as a Multiple-Input/Multiple-Output (MIMO) Intelligent
Multi-Media Signal Processor (IMMSP). The input-output map f of such a MIMO
IMMSP resides in an appropriate Reproducing Kernel Hilbert Space F of nonlinear
functionals on the multi-media input space. The realization of the optimal
model is obtained by an appropriate orthogonal projection in F, subject to
design-specification and exemplary input-output-data constraints. Such optimal
realizations of MIMO IMMSPs naturally appear in the form of fuzzy neural
systems, the fuzziness of which is encapsulated in the reproducing kernel of
the space F that gave birth to them.
As required in most applications, the above approach will permit a MIMO IMMSP to be: structurally, a large-scale, parallel, distributed, nonlinear, time-varying, robust (capable of graceful degradation) and affordable system; and functionally, possessing one or more of the capabilities of adaptation, learning (with and without supervision) evolution, and discovery. These concepts and related developments will be illustrated by examples.
ABOUT THE SPEAKER:
Rui J.P. de Figueiredo, Ph.D., earned a
B.S. and M.S. from M.I.T., and a Ph.D. from Harvard. He is Professor (Above Scale) of Electrical
Engineering and Computer Science, Biomedical Engineering, and Mathematics, and
Director of the Laboratory for Intelligent
Signal Processing and Communications at UC Irvine. Professor de Figueiredo’s accomplishments and
honors can be found on page: http://www.lminsc.uci.edu/6honors.pdf