|
IJCNN'04
IJCNN'97
Favorites |
IEEE Neural Networks Society
Hawaiian Hilton Village Honolulu, Hawaii May 12, 2002
TUTORIAL SCHEDULE AND ABSTRACTS 8:00 AM - 10:00 AM T01 "An Introduction to Evolutionary Computation" by Hans-Paul Schwefel (T01 Abstract/Vita ) T02 "An Introduction to Neural Networks" by Jacek Zurada (T02 Abstract/Vita ) T03 "An Introduction to Fuzzy Systems" by Ramesh Krishnapuram (T03 Abstract/Vita ) T14 "Support Vector Machines" by M. Palaniswami (T14 Abstract/Vita )
10:15 AM to 12:15 PM
T04 "Particle Swarm Optimization" by Russ Eberhart and James Kennedy ( T04 Abstract/Vita ) T10 "Neural Nets For Diagnostics, Prediction and Control: Capabilities & Myths" by Paul Werbos (T10 Abstract/Vita ) T18 "An Introduction to Biological Sequence Analysis" by Mitra Basu (T18 Abstract/Vita ) T23 “Evolutionary Algorithms for Program Induction” by Peter J. Angeline (T23 Abstract/Vita ) T20 "Computational Intelligence for Data Mining" by Wlodzislaw Duch (T20 Abstract/Vita )
12:30 to 2:30 PM T05 "Artificial Immune Systems" by Dipankar Dasgupta (T05 Abstract/Vita ) T06 "Computational Intelligence for Scheduling" by David Corne (T06 Abstract/Vita ) T09 "The Legal Aspects of Computational Intelligence" by Dennis Fernandez (T09 Abstract/Vita ) T11 "Attention and Selection" by Ernst Niebur (T11 Abstract/Vita ) T19 "Intelligent Multimedia Processing" by Ling Guan (T19 Abstract/Vita ) T17 "Fuzzy Systems in Biomedicine" by Peter Adlassnig (T17 Abstract/Vita )
2:45 PM to 4:45 PM T08 "Artificial Life Systems" by Claus Wilke (T08 Abstract/Vita ) T21 "Neural Networks for Scene Analysis" by DeLiang Wang (T21 Abstract/Vita ) T22 "Multi-Criterion Evolutionary Algorithms" by Kalyanmoy Deb (T22 Abstract/Vita ) T13 "An Introduction to Cultural Algorithms" by Robert G. Reynolds (T13 Abstract/Vita ) T16 "Neural Network Models for Speech and Image Processing" by B. Yegnanarayana (T16 Abstract/Vita ) T12 "Type II Fuzzy Logic" by Jerry Mendel (T12 Abstract/Vita )
WCCI 2002 General Chair: David Fogel Tutorial Chair: Mary Lou Padgett m.padgett AT ieee.org _______________________________________________________________________________
An Introduction to Evolutionary Computation
Hans-Paul Schwefel
schwefel@evol.cs.uni-dortmund.de
ABSTRACT
We start with embedding Evolutionary Algorithms (EA) into the broader class of Computational Intelligence (CI) methods. We emphasize their role as models of organic adaptation as well as tools for difficult optimization tasks - without ignoring classical optimization.
Next, we draw a general frame of the EA basics: variation and selection or exploration and exploitation. A short look onto the history of EA reveals some differences between Evolutionary Programming (EP), Genetic Algorithms (GA), and Evolution Strategies (ES). This is somehow related to search spaces of different type, i.e. boolean, integer/discrete, continuous, combinatorial. Here, we go into some details, so that practitioners can apply what they learn immediately thereafter.
A brief summary of theoretical results concerning global convergence and convergence rates will be given for non-theoreticians. This will destroy some wrong beliefs as well as euphemistic hopes.
Finally, some applications will be mentioned, among them experimental amelioration and numerical multi-objective optimization examples
VITA
Hans-Paul Schwefel studied Aero- and Space-Technology at the Technical University of Berlin (TUB). Before and after receiving his Diploma in 1965 he worked at the Hermann-Foettinger-Institute of Hydrodynamics, from 1967 to 1970 at the industrial AEG research institute, and from 1971 to 1975 again at the TUB, from where he got his Dr.-Ing. degree in 1975. Coherent during that period at Berlin was the development of a new experimental and later on also numerical optimization method called Evolution Strategy. From 1976 to 1985 he acted as senior research fellow at the Research Centre (KFA) Juelich, where he was head of a Computer Aided Planning Tools group. Since 1985 he is holder of a chair for Systems Analysis at the University of Dortmund, Department of Computer Science. From 1990 to 1992 he acted as dean of the faculty, from 1998 to 2000 as prorector for research at the university. He is member of the editorial boards of the journals Evolutionary Computation, IEEE Transactions on Evolutionary Computation, BioSystems, and Natural Compting, and in 1990 he was co-founder of the conference series "Parallel Problem Solving from Nature (PPSN)."
Hans-Paul Schwefel Chair of Systems Analysis Dept. of Computer Science University of Dortmund D-44221 Dortmund Germany schwefel@ls11.informatik.uni-dortmund.de
Hans-Paul Schwefel ---------------------------------------------------------------------------- Hans-Paul UNIVERSITAeT DORTMUND telefax: 49 231 9700 959 ///// Schwefel Dept. of Computer Science voice: 49 231 9700 952 ____UNI DO Prof.Dr. Chair of Systems Analysis secretary:49 231 9700 951 \\\*\//// D-44221 Dortmund, Germany \\\\\// NEW e-mail: hps@uDo.edu \\\\/ www: http://Ls11-www.cs.uni-dortmund.de ------------------------------------------------------------------- also: INFORMATIK CENTRUM DORTMUND (ICD) Center for Applied Systems Analysis (CASA) Joseph-von-Fraunhofer-Str. 20 D-44227 Dortmund phone/fax as above e-mail: hps@icd.de www: http://www.icd.de/ICD/CASA/ICD-CASA.html ----------------------------------------------------------------------------
An Introduction to Neural Networks
Jacek Zurada
jacek@comp.nus.edu.sg
ABSTRACT:
This tutorial reviews basic feedforward and recurrent architectures of neural networks. Overview of unsupervised and supervised learning paradigms is discussed in context of a single neuron and of selected architectures' learning (Hebbian, Perceptron, Delta, Winner-Take-All, and Correlation and Error Back Propagation learning). This tutorial discussion is richly illustrated with figures, and key learning algorithms are provided in closed forms. Also, links to useful NN modeling software are provided to assist attendees in future modeling efforts.
VITA Dr. Jacek M. Zurada is the S.T. Fife Alumni Professor of Electrical and Computer Engineering at the University of Louisville, Louisville, Kentucky. He has authored or co-authored more than 160 publications in neural networks and signals and systems area. He has authored a pioneering academic neural networks textbook (Introduction to Artificial Neural Systems), and co-edited two other volumes in 1994 and 2000 (Computational Intelligence: Imitating Life, and the Knowledge Based Neurocomputing). Since 1998 Dr. Zurada has been the Editor-in-Chief of IEEE Transactions on Neural Networks. He has also been an invited plenary speaker at many national and international conferences. He is IEEE Fellow and IEEE Distinguished Speaker.
An Introduction to Fuzzy Systems
Raghu Krishnapuram
kraghura@in.ibm.com
ABSTRACT
The tutorial will cover the basics of fuzzy set theory including membership functions, operations on fuzzy sets (fuzzy aggregation operators), extension principle, linguistic variables and hedges, fuzzy logic, fuzzy systems, and fuzzy clustering
VITA Raghu Krishnapuram received his Ph.D. degree in electrical and computer engineering from Carnegie Mellon University, Pittsburgh, in 1987. From 1987 to 1997, he was on the faculty of the Department of Computer Engineering and Computer Science at the University of Missouri, Columbia. In 1997, Dr. Krishnapuram joined the Department of Mathematical and Computer Sciences at the Colorado School of Mines (CSM), Golden, Colorado, as Full Professor. Currently he is at IBM India Rearch Lab, New Delhi. Prof. Krishnapuram's research encompasses many aspects of fuzzy set theory, neural networks, pattern recognition, computer vision and image processing. He has published over 150 papers in journals and conferences in these areas. He is currently an associate editor of the IEEE Transactions on Fuzzy systems and he is a co-author (with J. Bezdek, J. Keller and N. Pal) of the book "Fuzzy Models and Algorithms for Pattern Recognition and Image Processing". His current research interests include Web mining and e-commerce.
Particle Swarm Optimization
Russell Eberhart and Jim Kennedy
eberhart@tech.iupui.edu
ABSTRACT
Particle swarm optimization (PSO) is an evolutionary computation technique that uses a population of potential solutions. It is unlike other evolutionary algorithms (EAs), however, in that each potential solution is assigned a random velocity, and the potential solutions, called particles, are then "flown" through the problem space. The basic algorithm is very simple, comprising two lines of computer code; the speed of convergence is typically faster than other EAs on benchmark problems. The tutorial will address the artificial life roots of PSO, its variations, and illustrate PSO's usefulness with example applications. Participants will be provided with source and executable code. The tutorial presenters are co-authors of the book, "Swarm Intelligence," published in 2001
VITA Russell C. Eberhart is the Chair and Professor of Electrical and Computer Engineering at the Purdue School of Engineering and Technology, Indiana University Purdue University Indianapolis (IUPUI). He received his Ph.D. from Kansas State University in electrical engineering. He is co-editor of a book on neural networks, and co-author of Computational Intelligence PC Tools, published in 1996 by Academic Press. He is co-author of a book with Jim Kennedy and Yuhui Shi entitled Swarm Intelligence, published by Morgan Kaufmann/Academic Press in April 2001. He is Associate Editor of the IEEE Transactions on Evolutionary Computation. He was awarded the IEEE Third Millenium Medal. In January 2001, he became a Fellow of the IEEE. He is co-author with Yuhui Shi of a book entitled Computational Intelligence: Concepts to Implementations, to be published in late 2002.
Artificial Immune Systems
Dipankar Dasgupta
ddasgupt@memphis.edu
ABSTRACT
The immune system is an intricate network of specialized tissues, organs, cells, and chemicals. Its function is to distinguish entities within the body as "self" or "nonself" and to eliminate or neutralize those that are nonself (or dangerous). To accomplish its tasks, the immune system has evolved two mechanisms: innate (nonspecific) immunity and adaptive (specific) immunity, which are linked to and influence each other. The system uses learning, memory, and associative retrieval to solve recognition and classification tasks. These remarkable information-processing abilities of the immune system provide several important aspects in the field of computation. Like other biologically motivated approaches (Neural Networks, Genetic Algorithms, etc.), the Artificial Immune System is also a rapidly emerging field. Artificial Immune Systems are used in pattern recognition, fault detection, computer security, and a variety of other applications in science and engineering. The tutorial will discuss different immunological mechanisms and their relation to information processing and problem solving.
VITA Dr. Dipankar Dasgupta is a faculty of Computer Science in Mathematical Sciences department at the University of Memphis, Tennessee. His research interests are broadly in the area of scientific computing, tracking real-world problems through interdisciplinary cooperation. He has been working on computational models of the immune system and their application for several years. He published more than 65 papers in book chapters, journals, and international conferences. He edited a book "Artificial Immune Systems and Their Applications" published by Springer-Verlag, 1999. Dr. Dasgupta is a member of IEEE, ACM and regularly serves as program committee member in many International Conferences. He organized special tracks and workshops on Artificial Immune Systems and offered tutorials on the topics at International Conferences since 1997. Dr. Dasgupta currently editing a special issue (on Artificial Immune Systems) of IEEE Evolutionary Computation Journal.
Computational Intelligence for Scheduling
David W. Corne
D.W.Corne@rdg.ac.uk
ABSTRACT
About 20% of this tutorial will attempt to introduce scheduling -- a class of problems found commonly in manufacturing, project management and countless other areas of industrial and commercial importance -- to computational intelligence scientists. This will be done with a view towards serving such scientists with a way forward in applying their knowledge to this area. The remaining bulk of the tutorial will survey (although in a rather eclectic fashion) how evolutionary computation (mainly) and other computational intelligence technologies have been and are being applied to scheduling problems. In doing so I will attempt to pull out aspects of best practice, pitfalls, and emerging methods and opportunities
VITA David Corne is a Reader in Evolutionary Computation, and heads the Parallel, Emergent and Distributed Archiectures Laboratory at the University of Reading, England. He is an Associate Editor of the IEEE Transactions on Evolutionary Computation, was a Founding Editor of the Journal of Scheduling, and has various other editorial board memberships. His research interests are in aspects and applications of computational technologies, particularlt evolutionary computation, and applied particularly in the areas of bioinformatics, scheduling, and data mining.
Artificial Life systems
Claus O. Wilke
wilke@caltech.edu
ABSTRACT
In genetic algorithms and evolution strategies, one usually studies the dynamics of evolution in the context of a given external fitness function. In artificial life, on the other hand, one is typically more interested in the dynamics of simulated ecosystems, in which fitness is an implicit function of the state of the ecosystem at a given time. Research with artificial life forms is geared towards the development of a general theory of evolutionary processes. By comparing results obtained with such artificial life forms to those obtained from carbon-based life forms, one can identify general principles of living systems that are independent of the particular substrate from which they are built.
The tutorial consists of two parts. In the first part, an overview is given over the history of artificial life systems from the early 1970's till today [1-4]. The main emphasis is laid on digital organisms, which are computer programs that self-replicate, mutate and evolve. In the second part, the currently most widely used simulation system for digital organisms, the Avida platform [4], is introduced. Avida, which is being developed and employed at Caltech and Michigan State University, allows for sophisticated experiments with digital organisms. The second part of the tutorial covers both the design of Avida and some recent experiments in which Avida has been used to investigate fundamental principles of evolving systems [5-8].
[1] M. Conrad and H. H. Pattee. Evolution experiments with an artificial ecosystem. J. theoret. Biol. 28, 393-409 (1970).
[2] S. Rasmussen, C. Knudsen, R. Feldberg, and M. Hindsholm. The Coreworld: Emergence and evolution of cooperative structures in a computational chemistry. Physica D 42, 111-134 (1990).
[3] T. Ray. An approach to the synthesis of life. In "Artificial Life II", C. G. Langton et al. (Addison-Wesley 1992), p. 371-408.
[4] C. Adami. Introduction to Artificial Life (Springer, New York, 1998).
[5] R. E. Lenski, C. Ofria, T. C. Collier, and C. Adami. Genome complexity, robustness and genetic interactions in digital organisms. Nature 400, 661-664 (1999).
[6] C. Adami, C. Ofria, and T. C. Collier. Evolution of biological complexity. Proc. Natl. Acad. Sci. USA 97, 4463-4468 (2000).
[7] C. O. Wilke, J. L. Wang, C. Ofria, R. E. Lenski, and C. Adami. Evolution of digital organisms at high mutation rates leads to survival of the flattest. Nature 412, 331-331 (2001).
[8] C. O. Wilke and C. Adami. Interaction between directional epistasis and average mutational effects. Proc. R. Soc. Lond. B 268, 1469-1474 (2001).
VITA Claus Wilke was born 1972 in Hattingen, Germany. He graduated with a Diploma in high-energy physics from the University of Bochum, Germany in 1996, and received his PhD in physics also from the University of Bochum in 1999. His PhD thesis was on the theoretical modeling of evolutionary dynamics in time-dependent environments. After a short three-month stay as a postdoc at the University of Luebeck, Claus Wilke moved to Pasadena, California, to work as a postdoc at Caltech with Chris Adami, a position which he still holds to this day. Claus Wilke is currently mainly interested in the theoretical modeling of populations, quasispecies effects, and the dynamics of self-replicating computer programs (digital organisms).
The Legal Aspects of Computational Intelligence
Dennis Fernandez
iploft@iploft.com
ABSTRACT
Intellectual Property Rights protection for various computational intelligence technologies will be discussed, particularly focusing on effective patent strategies to protect novel processing apparatus and methods as well as software algorithms and data structures.
VITA The Legal Aspects of Computational Intelligence
by Dennis Fernandez
BAR MEMBERSHIPS
U.S. Patent & Trademark Office (Reg. No. 34,160)
California; Massachusetts; Washington D.C.
EDUCATION
Northwestern University, B.S. Electrical Engineering, 1983
Suffolk University, J.D., 1989
EXPERIENCE
Venture Capitalist at Walden Group, and Vertex Management
General Counsel at NeoParadigm Labs (chip start-up)
Patent Attorney at Fenwick & West, and Nutter McClennen & Fish
Engineering positions at Digital Equipment, Raytheon, and Racal
ACTIVITES
Lecturer at University of California (Ext) and Stanford University
Editorial Advisory Boardmember for Red Herring Magazine
Boardmember for NASA Technology Incubator
Participated in industry speaking panels on patents and technology
Neural Networks: The Link Between, Vision, Consciousness and Practical Applications
Paul Werbos
pwerbos@nsf.gov
ABSTRACT
This tutorial will be an expanded version of a plenary talk given last year in Brazil and at an SMC conference, trying to provide a global vision of the neural network field aimed at the needs of engineering application, with links to the literature on consciousness. The neural net approach has led to new specific tools for prediction, diagnosis and intelligent control which are far more powerful than what is well known even within the neural net field. As an example, even the most sophisticated neurocontrol designs can scale much better to large-scale problems when the nonlinear mapping networks embedded within them have more power than traditional Hebbian or vanilla MLP networks. The tutorial will cover many common issues and pitfalls in neural prediction and control, including those discussed in the review paper posted at www.iamcm.org. By allowing time for interaction, it should allow greater depth in those topics of interest to the attendees, which could range anywhere from new efforts in aerospace control through to quantum information processing, all of which fit into this context. It will also be suggested that true "neuromorphic" engineering should try to account for fundamental principles necessary to brain-like capabilities which are not yet known to mainstream neuroscience
VITA Vitae
Paul J. Werbos holds four degrees from Harvard and the London School of Economics in: (1) economics; (2) internationalpolitical systems, emphasizing European economic institutions; (3) applied mathematics, with a major in quantum physics and a minor in decision and control; (4) applied mathematics, towards an interdisciplinary Ph.D. thesis. Prior to that, during high school, he obtained an FCC First Class Commercial Radiotelephone License, and took undergraduate and graduate mathematics courses at Princeton and the University of Pennsylvania.
For about four years after the PhD, he taught courses at Maryland in quantitative methods and global futures, and performed research in intelligent systems for policy application. Then for nine years he worked at the Department of Energy evaluating and developing a wide range of energy forecasting models. In 1989 he joined NSF as a program director in the ECS Division with emphasis on Neuroengineering. He also initiated the SBIR topic 26 which emphasizes fuel-cell automobiles, for which he is Technical Coordinator. Within the Knowledge Modeling and Computational Intelligence (KMCI) area, his main goal is to maximize the development and dissemination of step-by-step advances in systems design which will lead to an understanding and replication of the general kind of learning-based intelligence.
He has served as President of the International Neural Network Society, where he is still on the Governing Board. He also serves on the AdCom of the IEEE Systems, Man and Cybernetic Society, and the Neural Networks Technical Committee of the IEEE Neural Networks Council.
RESEARCH INTERESTS
the development of learning designs, as per the previous paragraph links to neuroscience and efforts to understand the human mind more deeply issues in the foundations of quantum theory, particularly related to solitons and PDE issues involving sustainable growth, with partial emphasis on the role of advanced technologies.
Attention and Selection
Earnest Niebur
niebur@cbis.ece.drexel.edu
ABSTRACT
tThe tutorial focuses on functional mechanisms of the selection of vant information in neural information processing, with an emphasis on primates. Selective attention is nature's solution to a computational dilemma. On the one hand, a large variety of sensors in different modalities is required to monitor ongoing events in the environment. On the other hand, the amount of information provided by the multitude of sensors by far exceeds the information processing capacity of the brain. Likewise, a complete ``verbatim'' storage of all sensory information would also exhaust the available storage capacity within a short period of time. Therefore, appropriate filtering which eliminates all sensory input except for a carefully selected small subset is essential for the efficient functioning of biological information processors. This filtering process is commonly referred to as ``selective attention'' and the tutorial provides an overview of the functional and computational properties of this perceptive function in biological computation. Intelligent selection of relevant information will also be of crucial importance in technical information processing systems once they reach a certain level of complexity and the availability of raw sensory information is secured. The primate nervous system may provide a powerful model and the inspiration to develop strategies for intelligent information selection in such systems.
references:
Steinmetz, P. N. and Roy, A. and Fitzgerald, P. and Hsiao, S. S. and Johnson, K. O. and Niebur, E., "Attention Modulates Synchronized Neuronal Firing in Primate Somatosensory Cortex", Nature, vol 404, pp 187-190, 2000
Niebur, E. and Koch, C, "Computational Architectures For Attention", in "The Attentive Brain", MIT Press, R. Parasuraman (ed.). Chapter 9, pp 163-186, Cambridge, MA 1998.
Itti, L. and Niebur, E. and Koch, C. "A model of saliency-based fast visual attention for rapid scene analysis", IEEE Transactions on Pattern Matching and Machine Intelligence, vol 20, pp 1254-1259, 1998
Niebur, E. and Koch, C. "A model for the neuronal implementation of selective visual attention based on temporal correlation among neuronsz", Journal of Computational Neuroscience, vol 1(1), pp 141-158, 1994.
Niebur, E. and Koch, C. and Rosin, C. "An oscillation-based model for the neural basis of attention", Vision Research, vol 33, pp 2789-2802, 1993.
2001- Associate Professor, Zanvyl Krieger Mind/Brain Institute and Department of Neuroscience and Department of Psychological and Brain Sciences, The Johns Hopkins University, Baltimore, USA.
-- Ernst Niebur, PhD Krieger Mind/Brain Institute Assoc. Prof. of Neuroscience Johns Hopkins University niebur@jhu.edu 3400 N. Charles Street (410)516-8643, -8640 (secr), -8648 (fax), -3357 (lab) Baltimore, MD 21218
VITA Attention and Selection
by Ernst Niebur
Education
1982 Diplom Physiker (M.Sc.) University of Dortmund, W. Germany
1988 Dr es sciences (PhD), University of Lausanne, Switzerland
Professional Experience
1989-95 Research fellow (since 1993: Senior Research Fellow), Computation and Neural Systems program, California Institute of Technology, Pasadena, CA.
1995-2001 Assistant Professor, Zanvyl Krieger Mind/Brain Institute and Department of Neuroscience, The Johns Hopkins University, Baltimore, USA.
2001- Associate Professor, Zanvyl Krieger Mind/Brain Institute and Department of Neuroscience and Department of Psychological and Brain Sciences, The Johns Hopkins University, Baltimore, USA.
Type II Fuzzy Logic: Expanded and Enhanced Fuzzy Logic
Jerry M. Mendel
mendel@sipi.usc.edu
ABSTRACT:
Fuzzy logic systems (FLSs) are, as is well known, comprised of rules. Quite often, the knowledge that is used to construct these rules is uncertain. This leads to rules whose antecedents or consequents are uncertain, which translates into uncertain antecedent or consequent membership functions. Type-1 fuzzy logic systems, whose membership functions are type-1 fuzzy sets, are unable to directly handle-model-such rule uncertainties. Type-2 fuzzy logic systems-in which antecedent or consequent membership functions are type-2 fuzzy sets-can. Type-2 fuzzy sets are fuzzy sets whose membership functions are themselves type-1 fuzzy sets; they are very useful in circumstances where it is difficult to determine an exact membership function for a fuzzy set; hence, they are useful for incorporating uncertainties. Exactly what these uncertainties are will be covered in this tutorial
We may view computing the defuzzified output of a type-1 FLS as analogous (but not equal) to computing the mean of a probability density function. Just as variance provides a measure of dispersion about the mean, and is almost always used to capture more about probabilistic uncertainty in practical statistical-based designs, a FLS also needs some measure of dispersion to capture more about rule uncertainties than just a single number. A type-2 FLS provides this measure of dispersion, and that measure seems to be as fundamental to the design of systems that include linguistic uncertainties (that translate into rule uncertainties) as variance is to the mean.
In a probabilistic model, if the variance goes to zero, then there no longer is any dispersion about the mean, and, in essence, the mean now describes a deterministic quantity. In this way, we can include a deterministic model within a probabilistic model. In a type-2 FLS, if there are no rule uncertainties, then it reduces to a type-1 FLS. In this way, a type-1 FLS is included within a type-2 FLS.
Zadeh lately has been advocating "computing with words" [CW] and using fuzzy logic (FL) to do this. Usually, words mean different things to different people and so there is uncertainty associated with words, which means that FL must somehow use this uncertainty when it computes with words. Type 1 FL cannot do this, but type 2 FL can.
So, if we are to compute with words, or to handle rule uncertainties, we need to become knowledgeable about type-2 fuzzy logic and its application to a type-2 FLS. The purpose of this tutorial session is to provide attendees with up-to-date knowledge about type-2 FLSs. It will cover: motivation for type-2 fuzzy sets-the uncertainties present in a rule-based FLS, the language of type-2 fuzzy sets, operations on type-2 fuzzy sets, type-2 relations and compositions, the centroid of a type-2 fuzzy set, type-2 FLSs, the kinds of applications that seem to be most appropriate for type-2 FLSs, some applications, software, the future.
VITA At USC
Came to USC in 1974 from McDonnell Douglas Professor, Electrical Engineering-Systems Associate Director of Education, Integrated Media Systems Center (IMSC) Executive Director of Planning, Engineering Academic Center (EAC)
Education..
Ph. D. in Electrical Engineering, Polytechnic Institute of Brooklyn, Brooklyn, NY M. S. in Electrical Engineering, Polytechnic Institute of Brooklyn, Brooklyn, NY B. S. in Mechanical Engineering, Polytechnic Institute of Brooklyn, Brooklyn, NY
Honors and Awards..
Fellow of the IEEE Distinguished Member of the IEEE Control Systems Society Tau Beta Pi Pi Tau Sigma Sigma Xi SEG 1976 Outstanding Presentation Award for a paper on the application of Kalman Filtering to deconvolution, 1976 Best Transactions Paper Award for a paper on maximum-likelihood deconvolution in the IEEE Trans. on Geoscience and Remote Sensing, 1983 Signal Processing Society Paper Award for a paper on identification of nonminimum phase systems using higher-order statistics in the IEEE Trans. on Acoustics, Speech, and Signal Processing, 1992 Phi Kappa Phi book award for 1983 research monograph on seismic deconvolution Burlington Northern Faculty Achievement Award, 1995 IEEE Centenniel Medal, 1984 President of the IEEE Control Systems Society in 1986 Service Award from the School of Engineering at USC, 1993 IEEE Third Millenium Medal, 2000
Research Interests.
type-2 fuzzy logic: theory type-2 fuzzy logic systems: applications in signal processing, communications and control
Publications.
more than 380 technical papers author or editor of eight books, including Uncertain Rule-Based Fuzzy Logic Systems: Introduction and New Directions (Prentice-Hall, 2001) Lessons in Estimation Theory for Signal Processing, Communications and Control (Prentice-Hall, 1995) Maximum-Likelihood Deconvolution (Springer-Verlag, 1990) Two tutorials published in the IEEE Proceedings: "Fuzzy Logic Systems for Engineering," Proc. of IEEE, vol.83, pp. 345-377, March 1995. Corrections to this paper appear in vol. 83, p. 1293, Sept. 1995. "Tutorial on Higher-Order Statistics (Spectra) in Signal Processing and System Theory: Theoretical Results and Some Applications," Proc. of IEEE, vol. 79, pp. 278-305, 1991.
Designing Large-Scale Multi-Agent Systems Using Cultural Algorithms
Robert G. Reynolds
savathome@msn.com
ABSTRACT
Cultural Algorithms are computational models of Cultural Evolution. They consist of two basic components, a population space, and a belief space. The two components interact by means of a Vote-Inherit-Promote or VIP protocol. A variety of paradigms have been used to model the population component including Genetic Algorithms, Genetic Programming, Evolution Strategies, and Evolutionary Programming, Cellular Automata, and Multi-agent systems among others. Likewise the knowledge acquired by the problem solving activities of the population can be stored in the belief space in the form of production rules, semantic networks, version spaces, among others.
As such, Cultural Algorithms represent a general framework for producing hybrid evolutionary systems that integrate evolutionary search and symbolic reasoning together. Cultural Algorithms are particularly useful for problems whose solution requires extensive domain knowledge. The tutorial starts by providing a brief motivation for the Cultural Algorithm framework. Next the basic choices for configuring Cultural Algorithms are presented and the motivation for selected particular configurations will be discussed. The tutorial will conclude with example applications from a variety of application areas including function optimization, knowledge based system design and maintenance, software engineering, and the modeling of complex social systems.
VITA Dr. Robert G. Reynolds received his M.S. and Ph.D. in Computer Science from the University of Michigan-Ann Arbor in 1978 and 1979 respectively in Artificial Intelligence specializing in Genetic Algorithms. He is currently a Professor of Computer Science at Wayne State University and an Adjunct Research Associate at the University of Michigan Ann Arbor. Dr. Reynolds is particularly interested in evolutionary models of Cultural Systems. He developed Cultural Algorithms as a computational framework in which to model cultural evolution. His research has been supported by grants from both NSF and industry. Dr. Reynolds has co-authored two books, and written over 100 articles in the area of evolutionary computation. He is currently an Associate Editor of IEEE Transactions on Evolutionary Computation, International Journal on Tools for Artificial Intelligence, and the International Journal of Software Engineering and Knowledge Engineering. He has been active in the Evolutionary Programming Society and served in various conference committees including Co-Chair of the technical program for Evolutionary Programming in 1995, and 1997.
Support Vector Machines
M. Palaniswami and D. Ralph
swami@postoffice.ee.mu.oz.au
ABSTRACT:
This tutorial is aimed at introducing engineers and Scientists to Support Vector Machines; to motivate and explain how SVMs are used for pattern classification, function approximation, and regression problems; and to demonstrate their potential both through simple examples and real world case studies.
Data classification, prediction, regression etc. are areas heavily used in information technology. Consider a bank when presented with a loan application for a commercial development: what are the best criteria to determine the likelihood of success, i.e. whether or not the loan should be approved? Another example comes from fisheries management: how can remote video sensing be used in the real-time identification of fish species in rivers?
Though these examples are rather different, they share a need for analysis of large amounts of data to interpret or predict properties of a very complex environment.
In recent years, Machine Learning has become a focal point in Artificial Intelligence. Support Vector Machines (SVMs) are a relatively new, general formulation for learning machines. SVMs perform exceptionally well on well pattern classification, function approximation, and regression problems.
Topics covered include a selected review of Machine Learning, fundamentals of SVMs, training and adaptive re-training techniques (optimization), incremental learning, comparison with Artificial Neural Nets, and various practical examples based on problems from loan default prediction(case studies), fish identification applications (case studies), and data and information fusion via density estimation.
VITA Marimuthu Palaniswami
Marimuthu Palaniswami obtained his B.E. (Hons) from the University of Madras , M.Eng. Sc. from the University of Melbourne , and Ph.D from the University of Newcastle , Australia . He is an Associate Professor at the University of Melbourne, Australia.
His research interests are in the fields of computational intelligence, nonlinear dynamics, computer vision, intelligent control and bio-medical engineering. He has published more than 150 conference and journal papers in these topics. He was an Associate Editor of the IEEE Tran. on Neural Networks and is on the editorial board of a few computing and electrical engineering journals. He served as a Technical Program Co-chair for the IEEE International Conference on Neural Networks, 1995 and has served on the programme committees of a number of international conferences including IEEE Workshops on Emerging Technologies and Factory Automation, Australian Conferences on Neural Networks, IEEE Australia-New Zealand Conferences on Intelligent Information Processing Systems. He has given several key note lectures and invited tutorials, in the areas of Machine Learning, Bio Medical engineering, and Control in the conferences such as International Joint Conference on Neural Networks, Como, 2000, International Joint Conference on Neural Networks, Washington 2001, International Conference on Bio Medical Engineering, 2001, Seventh Australia NewZealand Conference on Intelligent Information Processing Systems, Perth, 2001, International Conference on Artificial Intelligence, Tasmania 2000, International Conference on Control and Instrumentation, Chennai 2000. He has completed several industry sponsored projects for National Australia Bank, Broken Hill Propriety Limited, Defence Science and Technology Organisation, Integrated Control Systems Pty Ltd, and Signal Processing Associates Pty Ltd. He also received several ARCs, APA(I)s, ATERBS, DITARD and DIST grants for both fundamental and applied projects. He is also a recipient of foreign specialist award from the Ministry of Education, Japan .
Daniel Ralph
Daniel Ralph obtained his B.Sc. (Hons) from The University of Melbourne, Australia, and his M.S. and Ph.D. from the University of Wisconsin-Madison, USA . He is a senior academic at Cambridge University , UK after spending seven years at The University of Melbourne.
His research interests include algorithms for quadratic and nonlinear optimization problems, and their application to machine learning. Other research areas are discrete-time optimal control, model predictive control, and analysis of nondifferentiable systems. He has published 24 refereed papers in various journals and books, and co-authored a research monograph on an area of bilevel optimization called mathematical programming with equilibrium constraints. He is a member of the editorial board of SIAM J on Optimization and an associate editor both of Mathematics of Operations Research and J of the Australian Mathematical Society, Series B. His conference activities, apart from invited lectures and session organization, include chairing streams at the 1997 International Symposium on Mathematical Programming ( Lausanne ) and the 1998 International Conference on Nonlinear Programming and Variational Inequalities ( Hong Kong ). He has been the recipient of a number of research grants from the Australian Research Council including a current Large Grant on Support Vector Machines which is co-investigated by M Palaniswami and A Tsoi.
Neural Network Models for Speech and Image Processing
B.Yegnanarayana
yegna@iitm.ernet.in
ABSTRACT
Applications involving speech and images require extraction of information in the form of features from raw data, and then use those features for classification, storage and retrieval of information. Conventional methods of signal processing use linear methods or some simple nonlinear methods. But most of the time the information is embedded in features which require complex nonlinear processing of the data for extraction. Moreover, many classification models require nonlinear dividing surfaces in the feature space. Models based on artificial neural networks have been found to be very powerful for feature extraction and classification. This tutorial present basics of neural network models for features extraction and classification. In particular, neural network models will be discussed which can perform the following tasks: Extraction of higher order statistical features from data, capturing the distribution of the feature vectors, and combining evidence from several classifiers. Applications of these models for processing real speech and image data will be illustrated with demonstrations..
VITA Prof. Yegnanarayana is currently a professor in the Computer Science and Engineering department at IIT Madras, Chennai, India. He is with IIT Madras since 1980. Prior to joining IIT, he was a visiting associate professor of Computer Science at Carnegie-Mellon University in USA from 1977-1980. He was a member of faculty at the Indian Institute of Science, Bangalore from 1966 to 1978. He did BE, ME and PhD from IISc Bangalore, in 1964, 1966, and 1974, respectively. His research interests are in speech, image processing and neural networks. He has published several papers in these areas in IEEE and other international journals. He is also the author of the book "Artificial Neural Networks", published by Prentice-Hall of India, in 1999. He is a Fellow of the Indian National Academy of Engineering and a Fellow of the Indian National Science Academy
Fuzzy Systems in Biomedicine
Klaus-PeterAdlassnig
mes-office@akh-wien.ac.at
ABSTRACT
Fuzzy set theory and fuzzy logic have a number of characteristics that make them highly suitable for modeling uncertain, vague, and incomplete information, which concept-forming, patient state interpretation, and diagnostic and therapeutic decision-making are usually based upon. Linguistic medical entities such as symptoms, signs, test results, diseases and diagnoses, therapy proposals and prognostic information items including entities with temporal properties can be defined as fuzzy sets. The inherent vagueness of these entities will thus be conserved. Fuzzy logic offers reasoning methods capable of drawing strict as well as approximate conclusions. Medicine demands such a wide range of possibilities because the body of medical theory includes definitional, causal, statistical, and heuristic knowledge. Fuzzy automata can be used as high-level patient monitoring devices employing real time access to the various medical information systems; and fuzzy control achieves highly adaptive, gradual diagnostic or therapeutic decision output in various medical settings.
Content of the tutorial
This tutorial will include methods on how to construct fuzzy sets in medicine and how to process them in medical knowledge-based systems; furthermore fuzzy scores, fuzzy trend detection, fuzzy automata, and fuzzy control in medicine are presented.
It will demonstrate the applicability of fuzzy sets and fuzzy logic in medicine within a number of knowledge-based systems in the fields of internal medicine, intensive care medicine, and laboratory interpretive analysis. The clinical aims, applied methods, and achieved results will be presented.
Outline of topics:
1. Introduction
2. Construction of fuzzy sets
2.1. Expert definition
2.2. Supervised learning of fuzzy sets
3. Differential diagnostic consultation in internal medicine
3.1. Modeling of medical entities and medical relationships by fuzzy sets
3.2. Attach fuzziness to knowledge-based production rules
3.3. Fuzzy diagnostic consultation in rheumatology
4. Scores in medicine
4.1. Fuzzy scores
4.2. Score-based therapy entry criteria in patients with acute respiratory distress syndrome (ARDS)
5. Monitoring in intensive care medicine
5.1. Fuzzy trend detection
5.2. Fuzzy automata
6. Respirator control in intensive care medicine
6.1. Fuzzy control
6.2. Knowledge-based weaning from the respirator
7. Test interpretation in laboratory medicine
7.1. Modeling of temporal concepts by fuzzy sets
7.2. Application in toxoplasmosis serology test interpretation
8. Discussion
VITA Klaus-Peter Adlassnig received his M.S. degree in computer science from the Technical University of Dresden, Germany, in 1974. He joined the Department of Medical Computer Sciences of the University of Vienna Medical School, Austria, in 1976. In 1983, he obtained his Ph.D. degree in computer sciences from the Technical University of Vienna, Austria, with a dissertation on “A Computer-Assisted Medical Diagnostic System Using Fuzzy Subsets”.
Dr. Adlassnig was a postdoctoral research fellow with Professor Lotfi A. Zadeh at the Computer Science Division at the Department of Electrical Engineering and Computer Sciences of the University of California at Berkeley from 1984–86. He received his Venia docendi for Medical Informatics from the University of Vienna in 1988 and became Professor of Medical Informatics in 1992. In 1987, he received the Federal State Prize for excellent research in the area of rheumatology awarded by the Austrian Federal Ministry for Health and Environmental Protection. Since 1988, he has been head of the Section on Medical Expert and Knowledge-Based Systems at the Department of Medical Computer Sciences of the University of Vienna Medical School.
Prof. Adlassnig was a Visiting Professor at the Department of Medicine, Section on Medical Informatics, at the Stanford University Medical Center in summer 1993 and a Guest Professor at the Department of Electrical and Biomedical Engineering at the Technical University of Graz in 1998, 1999, 2000, and 2001. He spent the summer 2000 as a Visiting Scholar at the Department of Electrical Engineering and Computer Sciences, Computer Science Division, Berkeley Initiative in Soft Computing (BISC), University of California , Berkeley/U.S.A.
Since 2002, Prof. Adlassnig is the Editor-in-Chief of the International Journal “Artificial Intelligence in Medicine”, Elsevier Science Publishers B.V.
His research interests focus on computer applications in medicine, especially medical expert and knowledge-based systems and their integration in medical information and web-based health care systems. Prof. Adlassnig is highly interested in formal theories of uncertainty with emphasis on fuzzy set theory, fuzzy logic, fuzzy control, and related areas. He is equally interested in the theoretical and practical aspects of computer systems in medicine. Various aspects of the philosophy of science, especially the state and future impact of artificial intelligence, especially of medical artificial intelligence, are of high interest to Prof. Klaus-Peter Adlassnig.
An Introduction to Biological Sequence Analysis
Mitra Basu
eemb@engr.ccny.cuny.edu
ABSTRACT
Computational analysis of biological sequences is an emerging area at the forefront of biology and computer science. Efficient experimental techniques, primarily DNA sequencing, has led to an explosive growth of data. The need to interpret the data is becoming even more pressing. There is strong demand for sophisticated computational tools for a variety of tasks e.g., classifying sequences, detecting weak similarities, separating protein coding regions from noncoding regions in DNA sequences and predicting molecular structure and functions. Large databases of biological information create both challenging data mining problems and opportunities. Traditional computer science algorithms are unable to address many of the more interesting sequence analysis problems. This is due to the inherent complexity of biological systems and lack of a comprehensive theory of information processing at the molecular level. Machine learning approaches (e.g., neural network, hidden Markov model), on the other hand, are ideally suited for domains characterized by the presence of large amounts of noisy and incomplete data and the absence of general theories.
In this tutorial I will provide an overview of sequence analysis methods. I will include a discussion on DNA, RNA, amino acids, proteins and the transcription and the translation machinery that nature uses to process information from DNA to protein. Topics to be covered in this tutorial are : problem of sequence alignment, hidden Markov model, neural network, stochastic grammar. URL of useful internet resources and public databases will be made available.
VITA Mitra Basu received her Ph. D. in electrical engineering from Purdue University, West Lafayette, Indiana in 1985. The late Professor King-Sun Fu was her thesis supervisor. After completion of her doctoral work, she joined the faculty of electrical engineering department at the City College, where currently she is an associate professor. Dr. Basu's research interests include pattern recognition, learning systems and bioinformatics. She has joint appointments with the New York Center for Biomedical Engineering (http://www.ccny.cuny.edu/nycbe/) and the doctoral faculty of the Computer Science department. She is an associate editor for Pattern Recognition journal. For some of her recent publications see http://www-ee.engr.ccny.cuny.edu/faculty/basu.html.
Intelligent Multimedia Processing
LingGuan
lguan@ee.ryerson.ca
ABSTRACT
Human communication is intrinsically multimodal. With the advances of technology, modern communication systems will also become more and more multimodal. Hence, multimedia technologies represent new ground for research interactions among a variety of media such as speech, audio, image, video, text and graphics. Future multimedia technologies will need to handle information with an increasing level of intelligence, i.e., automatic recognition and interpretation of multimodal signals. This is particularly emphasized in MPEG-7 which focuses on the 'multimedia content description interface'. The description shall be associated with the content itself to facilitate fast and effective searching for all the media. Specifically, the MPEG-7 research domain will cover techniques for content-based indexing and retrieval: pattern recognition, face detection/recognition, and fusion of multimodality.
Intelligent multimedia processing shares three fundamental goals with biological systems: a) Universal data processing engine for multimodal signals; b) Multimodality; and c) Unsupervised clustering and/or supervised learning by examples. Because of these features, neural networks are attractive candidates for intelligent multimedia processing and recent activity in the area is a proof of this fact. The main attribute of neural computing is its adaptive learning capability which enables interpretations of possible variations of a same object or pattern, e.g., with respect to scale, orientation, and perspective. Moreover, they are able to accurately approximate unknown systems based on sparse sets of noisy data. Certain neural models also effectively incorporate statistical signal processing and optimization techniques. In addition, spatial/temporal neural structures and hierarchical models are promising for multirate, multiresolution multimedia processing. As a result, many successful applications of neural networks in intelligent multimedia processing, sometimes combined with fuzzy systems and evolutionary computation, have been reported.
Following an introduction on the state-of-the-art in intelligent multimedia processing, the following key technical aspects of intelligent multimedia processing will be presented and addressed in the tutorial: Neural networks and other computational intelligence models, learning paradigms, and architectures for multimedia processing, including RBF networks and SVM as the search engine in media retrieval, tree-structured SOM for automatic media retrieval over the Internet, Modular fuzzy neural networks for media fusion and dynamic resource allocation of video traffic, computational intelligence models for human-computer interface, general regression neural networks for feature selection, and evolutionary computation for optimizing some of the multimedia processing models and algorithms. In particular, we will focus the tutorial on the following areas:
1. Media indexing and retrieval. This is one of the most important areas in multimedia processing, considering that the success or failure of an multimedia application method primarily depends on the accuracy of the information extracted through indexing and retrieval. The popular search models used in retrieval are linear. Linear models cannot effectively simulate the non-linear nature of human perception. We introduce a non-linear projection based on a RBF network. By adjusting simple control parameters, the non-linear projection accurately captures the characteristics of human perception. The introduction of this model also eliminates rescaling, a very difficult, and formerly regarded as unavoidable task during the expansion of digital video libraries.
2.. Automatic retrieval. A well-designed media indexing and retrieval method is capable of projecting the low level features to the high level concepts. Recent research shows that, by incorporating relevance feedback and supervised-learning in the retrieval loop, the performance of a retrieval system can be substantially improved. However, this approach limits the application range of a retrieval system in Internet applications due its total dependence on human participation. We will present a new paradigm, which relies on a tree structure SOM and unsupervised learning. By extracting high-level features in a limited set of representative images, relevant images can be efficiently and automatically identified.
3. Dynamic resource allocation for video traffic. Researchers in industry and academia agree that network management plays the most critical role in delivering quality service to end users. One of the key issue is how to effectively and dynamically detect the resource requirement of each end user and appropriately allocate the user sufficient bandwidth. By appropriately selecting a set features characterizing the video traffic using the general regression neural network, the profile of video traffic can be effectively modeled and analyzed by a feedforward neural network or a modular fuzzy neural network. Evaluating the network link utilization on videos coded in MPEG-1 and MPEG-2 format shows the effectiveness of the method.
4. Human emotion/intention recognition. This is a research area with broad applications; from query model in retrieval to the development of bio-computers, to security/surveillance. We have developed a computational intelligence based method to tackle this problem. Audio/visual features extracted from human subjects are fed into a sequential selection algorithm/general regression neural network to analysis the effectiveness of the features. Due to the non-linear characteristics of the feature selection approach, we can demonstrate that the features selected are more effective in recognizing emotions than those selected by the linear methods such as the PCA. We then utilize modular fuzzy neural networks to fuse the audio visual media in identifying the emotions states of the human subjects.
Application examples and demos will be presented.
VITA Ling Guan received his Ph.D. Degree in Electrical Engineering from University of British Columbia, Canada in 1989. Until April 2001, he was on the Faculty of Engineering at the University of Sydney, Australia where was the founder and director of the Signal and Multimedia Processing Lab. Since May 2001, he has been a professor and Canada Research Chair at Ryerson Polytechnic University, Canada. In 1994, he was a visiting fellow at British Telecom. In 1999, he was awarded a fellowship by Australian Academy of Science/Japan Society for the Promotion of Science and a visiting professorial fellow at Tokyo Institute of Technology. In 2000, he was on sabbatical leave and a visiting professor at Princeton University.
His research interests include multimedia processing and systems, intelligent multimedia processing, optimal information search engine, signal processing for wireless multimedia communications, computational intelligence and machine learning, adaptive image and signal processing. He has been invited to present plenary speeches, tutorials, invited talks on these subjects at international conferences, government organizations, and research centers in the world, and has authored 150 technical publications in these areas. He is the editor/author of two books "Multimedia Image and Video Processing" and "Adaptive Image Processing: A Computational Intelligence Perspective."
He has been guest editors of numerous international journals, including a special issues on Intelligent Multimedia Processing for IEEE Transactions on Neural Networks, and a special issue on Computational Intelligence for Proceedings of the IEEE. He is an associate editor of IEEE Transactions on Evolutionary Computation, VLSI Signal Processing Systems for Signal, Image and Video Technology, and Journal of Electronic Imaging. He also serves on the editorial board of CRC Press' Book Series on Image Processing. He has been involved in organizing numerous international conferences. He played the leading role in the inauguration of the First IEEE Pacific-Rim Conference on Multimedia in Sydney, 2000, and served as the Founding General Chair.
Dr. Guan is a Senior Member of IEEE, and a member of IAPR and SPIE. He is currently serving on IEEE Signal Processing Society Technical Committee on Multimedia Signal Processing, and is a life-time member of the Advisory Board of IEEE Technical Committee on Neural Networks in Signal Processing
Computational Intelligence for Data Mining
Wlodzislaw Duch
duch@phys.uni.torun.pl
ABSTRACT
This tutorial presents computational intelligence approach to data mining, stressing the need for understanding of the data structure. At each step computer programs will be used in real-time on real-world examples to illustrate various procedures involved.
Forms of useful knowledge are discussed, including logical rules (crisp and fuzzy), decision trees, prototype-based rules and visualization techniques. The need for and advantages of various types of data analysis is explained.
A short description of the philosophy of integration of algorithms used in our GhostMiner software is presented, including an outline of 6 algorithms used in the software: IncNet, FSM, SSV, kNN, MDS and the committees of models.
Crisp and fuzzy logical rules are extracted from a few datasets. A tradeoff between accuracy/simplicity is explained using logical rules generated by FSM and SSV models.
A method for optimization of logical rules derived from these systems, exploring the tradeoff between rejection/error level, is presented.
A method to compute classification probabilities from any black-box system is presented. Assuming Gaussian uncertainties of measurements and crisp logical rules leads to analytical formulas that allow to optimize large complex sets of logical rules using gradient procedures. In effect interpretation is easy and accuracy is high.
Visualization of data and visualization of decision borders of classifiers is presented as an alternative method of data understanding; an example combining IncNet neural network and interactive multidimensional scaling (MDS) is presented.
Lessons from applications of this approach to a few real life problems are analyzed and simple logical rules for many datasets provided.
References: Duch W, Adamczak R, Grabczewski K, Methodology of extraction, optimization and application of crisp and fuzzy logical rules. IEEE Transactions on Neural Networks, 12 (2001) 277-306
VITA Wlodzislaw Duch is a professor of theoretical physics and applied computational sciences, since 1990 heading the Department of Informatics (formerly called a Department of Computer Methods) at Nicholas Copernicus University, Torun, Poland. His degrees include habilitation (D.Sc. 1987) in many body physics, Ph.D. in quantum chemistry (1980), and Master of Science diploma in physics (1977) at the Nicholas Copernicus University, Poland. He has held a number of academic positions at universities and scientific institutions all over the world. These include longer appointments at the University of Southern California in Los Angeles, and the Max-Planck-Institute of Astrophysics in Germany (every year since 1984), and shorter (up to 3 month) visits to the University of Florida in Gainesville; University of Alberta in Edmonton, Canada; Meiji University, Kyushu Institute of Technology and Rikkyo University in Japan; Louis Pasteur Universite in Strasbourg, France; King's College London in UK, to name only a few.
He has been an editor of a number of professional journals, including IEEE Transactions on Neural Networks, Computer Physics Communications, Int. Journal of Transpersonal Studies and a head scientific editor of the "Kognitywistyka" (Cognitive Science) journal. He has worked as an expert for the European Union science programs and for other international bodies. He has published 4 books and over 250 scientific and popular articles in many journals. He has been awarded a number of grants by Polish state agencies, foreign committees as well as European Union institutions.
Neural Networks for Scene Analysis
DeLiang Wang
dwang@cis.ohio-state.edu
ABSTRACT
A remarkable achievement of the perceptual system is scene analysis, which involves two basic processes: the segmentation of a scene into a set of coherent patterns and the recognition of memorized ones. Pattern recognition has been extensively studied in neural networks and is covered by many tutorials and standard textbooks. In this tutorial, I focus on scene segmentation. Closely related to scene segmentation are figure-ground segregation, which emphasizes segmentation of one object from the rest of the scene (background), and perceptual organization. I will first review standard neural network approaches to scene analysis, including the use of Boltzmann machines for figure-ground separation, feature-boundary systems for integrating region and boundary analyses, and classification-based approaches using Kohonen maps, multilayer perceptrons, and so on. I will then turn to oscillation-based approaches, introducing the oscillatory correlation theory and its mathematical and biological foundations. The application of the oscillatory correlation theory to a number of real world problems will be reviewed, including both visual and auditory scene analysis. Specific applications will be highlighted, and they include aerial and medical image segmentation, analysis of texture and motion images, speech segregation, and sound analysis based pitch and location.
VITA DeLiang Wang received the B.S. degree in 1983 and the M.S. degree in 1986 from Peking (Beijing) University, China, and the Ph.D. degree in 1991 from the University of Southern California, all in computer science. Since 1991, he has been with the Department of Computer and Information Science and the Center for Cognitive Science at the Ohio State University (OSU), where he is Professor. He was a visiting scientist in the Visual Sciences Laboratory of Harvard University in 1998-1999. His research interests include neural networks for perception, neurodynamics, neuroengineering, and computational neuroscience. He has published more than 40 papers in leading scientific journals, and numerous conference papers and book chapters. He is a recipient of the U.S. National Science Foundation Research Initiation Award in 1992 and the Office of Naval Research Young Investigator Award in 1996.
He currently serves on the editorial boards of IEEE Transactions on Neural Networks, Neurocomputing, Neurocomputing & Applications, and the Handbook of Brain Theory and Neural Networks (2nd Ed.). He has served either as organizer for or on program committee of many scientific conferences, including International Joint Conference on Neural Networks (IJCNN) and will serve as Program Co-Chair for IJCNN'03.
Multi-Objective Evolutionary Algorithms
Kalyanmoy Deb
K.Deb@cs.bham.ac.uk
ABSTRACT
Most real-world search and optimization problems are better posed as multi-objective optimization problems involving conflicting objectives. However, due to the lack of efficient solution techniques these problems are usually solved as single-objective optimization problems. Optimizing a product for minimum cost as well as for maximum reliability, for example, is often desired but is difficult to achieve with classical approaches. In reality such conflicting objectives give rise to a set of optimal solutions, known as Pareto-optimal solutions. In tackling such problems, a less-subjective and a useful approach is to first find a set of trade-off solutions and then to choose one from the set. In this tutorial, a number of state-of-the-art multi-objective evolutionary algorithms, capable of finding multiple trade-off solutions, will be discussed. A few case studies from different areas of science and engineering will demonstrate the usefulness of MOEAs in real-world applications. Finally, some salient challenges to the research and application of this emerging field will be discussed, thereby making the tutorial beneficial to researchers and practitioners alike
VITA Kalyanmoy Deb is the author of the first-ever book on the topic entitled `Multi-Objective Optimization Using Evolutionary Algorithms' (Chichester: Wiley, June 2001, 506 pages).
Evolutionary Algorithms for Program Induction
Peter J. Angeline
angeline@natural-selection.com
ABSTRACT
When the goal of a process is to induce a model for an observed or desired behavior, a natural inspiration is to model the behavior as a computer program. The ubiquity and familiarity of programs as explicit codified specifications of behavior make them an obvious and potentially human friendly representational choice. In addition, they offer the possibility of directly capturing optimal behaviors for embedded systems, autonomous vehicles, multi-agent systems, and other current engineering applications. Genetic programming, first described and popularized by John Koza, offers one method for transducing behaviors into an executable symbolic structure, or program. This tutorial will present an introduction to Genetic Programming and other methods for evolving executable structures. Issues involving program growth, representation, interpretation and manipulation will be covered in detail. Major extensions to the basic methods presented, such as inducing modular programs, incorporating memory, and multiple interacting programs, will be reviewed as well. Examples of applications including time series prediction, cooperative multi-agent systems, and other current engineering topics will be provided.
VITA
Dr. Peter Angeline received his Ph.D. from The Ohio State University in 1993. He has applied artificial and computational intelligence techniques to various contracts while working for IBM, Loral, Lockheed Martin, and 3M. He is currently a Senior Staff Scientist at Natural Selection, Inc. Dr. Angeline has edited 2 books on genetic programming, and served as a co-editor of a special issue of the journal Evolutionary Computation devoted to evolutionary program induction. Dr. Angeline currently sits on the boards of the Journal of Experimental and Theoretical Artificial Intelligence, Evolutionary Computation, and Genetic Programming and Evolvable Machines, and was the general chairman of the first Congress on Evolutionary Computation, held in 1999.
(top)This page (http://www.mlpadgett.org/index.html)
was most recently modified on January 14, 2006 and is maintained by
m.padgett(at)ieee.org. |