IJCNN 2005 Workshop

Achieving Functional Integration of Diverse Neural Models

http://openmap.bbn.com/~thussain/NeuralIntegration/index.html

Workshop will immediately follow IJCNN 2005
Hilton Bonaventure Hotel
Montreal, Quebec, Canada
August 4, 2005
7 - 10 pm

Workshop Program

Speaker Bios

Marc de Kamps studied physics at the University of Amsterdam. After obtaining an MSc in theoretical physics, he moved to experimental high energy physics. In the ZEUS experiment in the DESY accelerator, he was part of the offline reconstruction group, which sparked his interest in software engineering techniques. After obtaining his PhD, he moved to Frank van der Velde's group at the psychology department of Leiden University, where he implemented neural models of object-based attention in vision and natural language representation. Currently he works at the 'Robotics and Embedded Systems' Chair of the Technische Universitt Mnchen and he is deputy coordinator of nEUro-IT.net, an EU funded network in the area of NeuroIT, which organizes workshops, summer schools and also drafts a Roadmap for NeuroIT. He is currently working on a fast implementation of Fokker-Planck equations for leaky-integrate-and-fire neurons in a framework, with the aim of constructing biologically plausible neural networks that bridge the scale between spiking neurons and artificial neural networks.

Alexander Iversen has a Computer Engineering degree from Aalesund University College, Norway, and a BSc Honours degree in Computer Science from Heriot-Watt University, Edinburgh, UK. Mr Iversen is currently a PhD candidate at Heriot-Watt with interests in neural network techniques for pattern recognition. In this work he is also collaborating with the Norwegian Defence Research Establishment for applications in communications signal recognition.

Dr. Victor Eliashberg is a system engineer with a broad hands-on experience in the design and implementation of different types of digital, analog, and mixed-signal systems (hardware and software). He is a consulting professor at the Department of Electrical Engineering of Stanford University, and the president of Avel Electronics a private consulting company in Palo Alto. He has an eclectic background that includes control theory, electronics, computer science, mathematics, physics, neurobiology, and psychology and has been interested in the problem of a brain-like universal learning computer since the late sixties. As a system engineer, he is particularly concerned with the problem of system integration.

Xiaodong Gu was born in Nantong, Jiangsu Province, China, in 1970. He received the M.S. degree in communication and information system from Soochow (Suzhou) University, Suzhou, China, in 2000 and Ph.D. degree in signal and information processing from Peking University, Beijing, China, in 2003. Currently he is a Postdoctoral Fellow of Electronic Science and Technology Postdoctoral Research Station, with the Department of Electronic Engineering, Fudan University, Shanghai, China. He is taking charge of 2 projects about neural networks supported by China Postdoctoral Science Foundation and Shanghai Postdoctoral Science Foundation, and has taken charge of 1 sub-project supported by China National 863 Foundation. He published more than 30 papers in journals and conference proceedings as the first author. His current research interests include artificial neural networks, image processing, and pattern recognition. Dr. Gu serves as a referee for IEEE Transaction on Systems, Man, and Cybernetics, Part B. He was awarded 2001 NOKIA Scholarship by Peking University. He was the recipient of 2003 Excellent Graduate of Peking University, and was the recipient of 2005 Excellent Dissertation Award of Peking University. He was received Excellent Invited Report Awards in 12th and 13th China National Conference on Neural Networks in 2002 and 2003 respectively.

L. Andrew Coward was originally trained in theoretical physics at Cambridge and Lancaster Universities in the UK.. He worked for 30 years with Nortel Networks in Canada and the USA on the design of telecommunications switches (Central Office and PBX). Responsibilities included, at different times, custom integrated circuit design, hardware design, software design, design capture and information management environments for hardware and software, system reliability and system architecture. In 1999 he moved to academia in Australia to explore how an understanding of the ways in which extreme complexity constrains the architecture of electronic systems could be applied to understanding biological brains and to designing complex learning systems. His current academic appointment is at the Australian National University in Canberra, although he still spends quite a lot of time in Vancouver, Canada. Since 1999 he has had about 20 refereed papers published, and a book A Systems Approach to the Brain: from Neurons to Consciousness is appearing in 2005. He holds US patent 6,363,420 on Method and system for heuristically designing and managing a network. His website is http://cs.anu.edu.au/~Andrew.Coward/

Fernando J. Von Zuben received his B.Sc. degree in Electrical Engineering in 1991. In 1993, he received his M.Sc. degree, and in 1996, his Ph.D. degree, both in Automation from the Faculty of Electrical and Computer Engineering, State University of Campinas, SP, Brazil. He is currently an Associate Professor at the Department of Computer Engineering and Industrial Automation, from the State University of Campinas, SP, Brazil. The main topics of his research are computational intelligence, with emphasis on artificial neural networks and bio-inspired computing, autonomous navigation, multivariate data analysis, control and identification of dynamic systems. He coordinates open ended research projects in these topics, tracking real-world problems through interdisciplinary cooperation, being the advisor of 16 Master and Ph.D. Theses already concluded, and being the author of more than 150 research papers and book chapters. Von Zuben is a member of the IEEE Institute of Electrical and Electronics Engineers, INNS International Neural Network Society, AAAI American Association for Artificial Intelligence, SBA Brazilian Automation Society, and SBC Brazilian Computer Society. He has also been a member of the Program Committee of leading conferences and has served as referee for a number of scientific journals.

Talib Hussain received his B.A. in Physics, B.Sc. in Cognitive Science, and his M.Sc. and Ph.D. in Computer Science, all from Queen's University (Kington, Ontario). He is currently a Senior Scientist at BBN Technologies (Cambridge, MA) with a broad interest in learning and training for both machines and humans. He specializes in evolutionary computation and neural network techniques, and has studied their use for robotic navigation, logistics planning, scheduling, network routing optimization, course-of-action prediction and market prediction. Recently, he was the principal investigator (PI) on the ARL-sponsored Advocates and Critics for Tactical Behaviors project, which studied the use of evolutionary computation for tactical navigation of a simulated unmanned ground vehicle. He was also the PI on the DARPA-sponsored Gorman's Gambit project, which studied design issues involved in training teamwork skills to a platoon of human soldiers using modern multi-player game technology.

Workshop goals

The field of neural networks presents a very rich variety of models that have been applied to many different problems.  However, successful application of neural networks to large-scale problems has been a general weakness.  Further, the development of complex neural systems that demonstrate significant cognitive capabilities currently seems beyond reach.

There is a critical need for new ideas and techniques for leveraging existing research by integrating current models in meaningful ways, with the goal of producing functional solutions to complex problems.  Once we understand how to effectively model not only the detailed processing of specific neural components, but also the rich variety of interactions that may occur between those neural components (and other non-neural ones), we may begin to realize systems that scale well and are cognitively robust.

The goals of this workshop are to explore what current integrative approaches and techniques show promise, and identify potential high-payoff areas for future research.  Relevant research areas may include, for example, modularity, evolutionary neural systems, hybrid systems that integrate biological processes (such as immune systems and hormonal systems) with neural models, neural growth mechanisms, neural system engineering techniques.

The outcome of this workshop is the identification of a set of key research topics on functional neural integration, together with an understanding of the key technical issues, limitations and benefits of each topic.  The creation of working groups to pursue these topics will be discussed.

The impact of this workshop will be to stimulate new discussion on a critical topic that we as a field ignore at our own risk.  New ways of approaching large-scale neural solutions must always be at the fore. The development of effective working groups will be important in generating interest and support from funding agencies and other customers to keep advancing the state-of-the-art.

Organizer