Unified Neural Network Designs: The Key to Large-Scale Applications and Understanding the Brain
Dr. Paul Werbos, National Science Foundation, Arlington, VA USA

Overview: Simple forms of artificial neural networks (ANNs) have become widely used in niche applications -- credit evaluation, pattern recognition, load forecasting, nonlinear function approximation, etc. For such applications, it is enough to know a few equations and use easy off-the-shelf software. But demanding applications require much more. For example, many people use simple neural adaptive control designs whose theoretical stability guarantees require unrealistic assumptions and do not provide good transient response, but more powerful methods (see presentations at ebrains.la.asu.edu/~nsfadp;Neural Networks IJCNN2003 issue) have proven they can keep physical electric power systems running under disturbances three times as large as ANY of the simpler methods can handle. This talk will discuss how the larger vision of the neural network field -- designing MODULAR general-purpose systems -- has been carried forward in recent years, allowing both (1) an ability to handle larger and larger types of challenges, in a deeply principled mathematical way; (2) the gradual development of designs which could begin to explain the FUNCTIONAL intelligence of the brain as a whole system, something which nonfunctional "computational models" cannot do.

Biography: Paul J. Werbos is best known as the original inventor of backpropagation, as part of his Harvard PhD thesis, which was reprinted in full in his book the Roots of Backpropagation, Wiley 1994, along with his classic 1990 tutorial on backpropagation through time for Proc IEEE. He was one of the three original two-year presidents of the International Neural Network Society, and winner of the IEEE Pioneer Award. He is Program Director for Control, Networks and Computational Intelligence at NSF, which actively seeks more proposals in this area. He has also been active in many cross-cutting funding initiatives; for example, he serves on the Working Group for energy storage and distribution of the interagency Climate Change Technology Program, and coordinated the NASA-NSF-EPRI solicitation on space solar power (NSF 02-098). He is also on the Planning Committee of the Millennium Project of the United Nations University (http://millennium-project.org), and has published a few papers on quantum foundations and technology (see arXiv.org, physics and nonlinear systems). He also has two degrees in economics from Harvard and the London School of Economics
 

Tuesday, November 4, 9:00 - 10:00 a.m. 

Dynamics of Respiratory Neural Networks During Maturaiton
Dr. Metin Akay, Darmouth College, Hanover, NH, USA

Overview: Neural engineering is an emerging discipline to understand the organizational principles and underlying mechanisms of the biology of neural systems and to study the behavior dynamics and complexity of neural systems in nature.

It coalesces the engineering including electronic and photonic technologies, computer science, physics, chemistry, mathematics with the molecular, systems, cellular, cognitive and behavioral neuroscience. Therefore, the neural engineering deals with many aspects of basic and clinical problems associated with neural dysfunction including the representation of sensory and motor information, the electrical stimulation of the neuromuscular system to control the muscle activation and movement, the analysis and visualization of complex neural systems at mutli-scale from the single-cell and to the system levels to understand the underlying mechanisms, electrical stimulation of the cochlea, the development of novel electronic and photonic devices for experimental probing, the simulation studies, the design and development of human-machine interface systems and artificial vision sensors and neural posthesis to restore and enhance the impaired sensory and motor systems and functions.

In this presentation, we will present the ongoing research activities at the Neural Engineering and Informatics Lab at Dartmouth. We will discuss our recent finding about the relative contributions of maturation to the dynamical behavior of respiration during ontogeny in the neonate. We define and quantify changes in the complexity of the respiratory neural network that accompany maturation in piglets using the approximate entropy method which provides a model independent measure of the complexity (irregularity) of the underlying mechanisms of the respiratory network.
 

Metin Akay, Associate Professor of Engineering, Psychology and Brain Sciences, and Computer Science at Dartmouth received his B.S. and M.S. in Electrical Engineering from the Bogazici University, Istanbul, Turkey in 1981 and 1984, respectively and a Ph.D. degree from Rutgers University in 1990.
Prof. Akay has played a key role in promoting the biomedical education in the world by writing several prestigious books and editing the IEEE Biomedical Engineering Book Series. He is a senior member of IEEE, a member of Eta Kappa, Sigma Xi, Tau Beta Pi, The American Heart Association, and The New York Academy of Science. He also serves on the advisory board of several international journals including the IEEE T-BME, IEEE T-ITIB, Smart Engineering Systems etc. and NIH Bioengineering partnership study session and several NSF review panels.
He is a recipient of the IEEE EMBS Career Service for his outstanding contributions to the advancement of the scientific stature and visibility of IEEE-EMBS and extraordinary dedication to the promotion of biomedical engineering education in the world. He also received the IEEE Engineering in Medicine and Biology Society Early Career Achievement Award 1997 for outstanding contributions in the detection of coronary artery disease, in understanding of early human development, and leadership and contributions in biomedical engineering education.
 

Wednesday, November 5, 9:00-10:00 a.m. 

Nonconvergent Neural Memories for Robust Encoding of Noisy Sensory Data
Dr. Robert Kozma, The University of Memphis, Memphis, TN USA

Overview: Conventional digital computers store information encoded in strings of binary digits. We propose an alternative approach of pattern-based computing, in which information is stored in the form of spatial patterns of amplitude modulation of an aperiodic oscillatory carrier wave. This method is based on the observation of Freeman and colleagues, that sensory information processing in the central nervous system is realized via collective oscillations of sparsely but globally interacting neuronal populations. Our approach includes as special cases other models, including deterministic cellular automata, such as Conway's Game of Life, Chua’s cellular neural networks, as well as thermodynamic models like the Ising model and Hopfield’s neural network arrays. We demonstrate the feasibility of spatial pattern encoding on a number of difficult classification problems. We describe how phase transitions can be generated in a noisy system and how phase transition become helpful in generating a robust memory. Issues related to hardware implementation of these principles in chip designs are also addressed.

Biography: ROBERT KOZMA is presently Professor of Computer Science at the Department of Mathematical Sciences, The University of Memphis, TN, which he joined in 2000. He has held joint/faculty appointments with UC Berkeley; the University of Otago, New Zealand; Tohoku University, Japan; Delft University of Technology, The Netherlands. He has earned his PhD at TU Delft (1992). Dr Kozma has over 20-years of expertise in the interdisciplinary research that encompasses neuroscience and cognitive science, computer science, artificial intelligence, computational physics and applied mathematics. He has published 3 books, about 50 articles in international journals and books, and over 60 papers in peer-reviewed conference proceedings. He is a Senior Member of IEEE, member of the Neural Network Technical Committee of the IEEE Neural Network Society, chair of the Neurodynamics Special Interest Group of the International Neural Network Society. He has been on the Program Committee of about 20 international conferences in the field of intelligent computation and soft computing. He is Program Co-Chair of IJCNN’04, July 2004, Budapest.