Co-evolutionary learning offers a very attractive learning paradigm where how well a learner (individual) does depends on a dynamic environment that includes other learners (individuals) in the same or co-evolving populations. As the case for other types of learning, generalisation is a key issue in co-evolutionary learning [1]. Although there have been experimental studies on the robustness [2], which is closely related to our notion of generalisation, of co-evolved learners, no rigorous theoretical framework exists for measuring generalisation quantitatively in co-evolutionary learning. This talk [1] will introduce such a rigorous theoretical framework for measuring generalisation, which is applicable to any type of co-evolutionary learning, based on statistical machine learning theories. Different definitions of generalisation are discussed. Specific case studies, using iterated prisoner's dilemma games as examples, will be presented. Under the vigorous theoretical framework, we are able to estimate generalisation in co-evolutionary learning within certain bounds. Such estimation represents a major step forward in measuring generalisation for co-evolutionary learning in practice.

References:

[1] S. Y. Chong, P. Tino and X. Yao, "Measuring Generalization Performance in Co-evolutionary Learning," IEEE Transactions on Evolutionary Computation, conditionally accepted in March 2007.

[2] P. Darwen and X. Yao, "On evolving robust strategies for iterated prisoner's dilemma,'' In Progress in Evolutionary Computation, Lecture Notes in Artificial Intelligence, Vol. 956, Springer-Verlag, Heidelberg, Germany, pp.276-292, 1995.

Biography

Xin Yao is a Professor (Chair) of Computer Science at the University of Birmingham, UK. He obtained his BSc from the University of Science and Technology of China (USTC) in Hefei, China, in 1982, MSc from the North China Institute of Computing Technology in Beijing, China, in 1985, and PhD from USTC in Hefei, China, in 1990.

He was a postdoctoral research fellow at the Australian National University (ANU) in Canberra in 1990-91 and at CSIRO Division of Building, Construction and Engineering in Melbourne in 1991-92. He was a lecturer, senior lecturer and an associate professor at the University College, the University of New South Wales (UNSW), the Australian Defence Force Academy
(ADFA) in Canberra in 1992-99. He took up a Chair of Computer Science at the University of Birmingham, UK, on the April Fool's Day in 1999.

Currently he is the Director of CERCIA (the Centre of Excellence for Research in Computational Intelligence and Applications) at the University of Birmingham, UK, a Distinguished Visiting Professor of the University of Science and Technology of China in Hefei, China, and a visiting professor of three other universities. He is an IEEE Fellow and a Distinguished Lecturer of IEEE Computational Intelligence Society. He won the 2001 IEEE Donald G. Fink Prize Paper Award and several other best paper awards. In his spare time, he does the voluntary work as the editor-in-chief of IEEE Transactions on Evolutionary Computation, an associate editor or editorial board member of several other journals, and the editor of the World Scientific book series on "Advances in Natural Computation". He has been invited to give more than 45 invited keynote and plenary speeches at conferences and workshops in 16 different countries. He is a Cheung Kong Scholar (Changjiang Chair Professor) of the Ministry of Education of the People's Republic of China.

His research has been well supported by research councils, government organisations and industry. His major research interests include evolutionary computation, neural network ensembles, and their applications. He has more than 230 refereed publications.

Top

Modern advancements in medicine have increased the opportunity for the application of evolutionary computation to problems such as cancer diagnosis, image analysis, chemotherapy scheduling, and drug discovery.

This is primarily due to the demonstrated utility of evolutionary algorithms for searching vast numbers of possible solutions and as an optimization method for pattern recognition models. The resulting models can have direct importance to human health. The keynote lecture will provide a review of the current approaches for discovery and pattern recognition in these problem areas, and will be of interest to both biologists and computer scientists. Future directions in personalized medicine will also be considered.

Biography

Gary B. Fogel currently serves as Vice President of Natural Selection, Inc. and specializes in applications of computational intelligence to problems in medicine and bio/chem-informatics. He joined Natural Selection, Inc. after completing the Ph.D. in biology from the University of California at Los Angeles in 1998 with a focus on the evolution and variability of histone proteins. While at UCLA, Dr. Fogel was a Fellow of the Center for the Study of Evolution and the Origin of Life and earned several teaching and research awards. His current efforts include the use of computational intelligence methods for cancer diagnostics, gene expression analysis, gene recognition, and small molecule activity prediction. Dr. Fogel is a senior member of the IEEE, member of Sigma Xi, the International Society for the Study of the Origin of Life, the Biomedical Engineering Society, and Evolutionary Programming Society. Dr. Fogel currently serves as an associate editor of BioSystems, IEEE Transactions on Evolutionary Computation, IEEE/ACM Transactions on Computational Biology and Bioinformatics, IEEE Computational Intelligence Magazine, and two other bioinformatics journals. In 2003 he co-edited "Evolutionary Computation in Bioinformatics," with David Corne and in 2007 he has co-edited "Computational Intelligence in Bioinformatics" (IEEE Press). Dr. Fogel served as technical co-chair for the 2001 Congress on Evolutionary Computation, program chair for the 2004 and 2006 Congress on Evolutionary Computation, and as general chair for the 2004 and 2005 IEEE Symposia on Computational Intelligence in Bioinformatics and Computational Biology. Dr.
Fogel is an elected member of the IEEE Computational Intelligence Society Administrative Committee.

Top

One of the key problems in AI is how an agent can best learn in a largely unsupervised manner via interactions with its environment.  Suitably designed games provide an excellent way to test approaches to this problem.  They provide competitive, dynamic, and unpredictable environments: a perfect domain for designing, testing and applying evolutionary algorithms and other computational intelligence paradigms.  They enable the emergence of open-ended intelligent behaviour and provide natural metrics to measure the success of that behaviour.

Two main ways to train agents given no prior export knowledge are temporal difference learning, and evolution (or co-evolution).  We’ll study ways in which these methods can train neural network based agents for games such as Othello, car racing, and Ms Pac-Man.  The results show that each method has important strengths and weaknesses, and understanding these leads to the development of new hybrid algorithms such as EvoTDL, where evolution is used to evolve a population of TD learners.  These hybrid algorithms have analogies in nature, in that they exhibit life-time learning (TDL), and species-level learning (Evolution).  They are also free to exploit less natural Lamarkian learning.   Examples will also be given of where seemingly innocuous changes to the learning environment have profound effects on the performance of each algorithm.

The main conclusion is that these are powerful methods capable of learning interesting agent behaviours, but there is still something of a black art in how best to apply them, and there is a great deal of scope for designing new learning algorithms.  The talk will also include live demonstrations of these learning algorithms.

Biography

Simon M. Lucas received his BSc. degree in Computer Systems Engineering from the University of Kent, UK, in 1986, and his PhD from the University of Southampton, UK, in 1991. He was appointed to a Lectureship at the University of Essex in 1992, and is currently a Reader in the Computer Science Department at Essex. His main research interests are evolutionary computation, pattern recognition, and using games as test-beds for and applications of computational intelligence. He is the inventor of the scanning n-tuple classifier, a super-fast and accurate OCR method, and is an Associate Editor of the IEEE Transactions on Evolutionary Computation.

He has played a significant role in organising many international conferences, and was Program Chair for IEEE CEC 2006, and has run many competitions associated with these conferences.  He is currently chair of the IEEE Computational Intelligence Society Games Technical Committee.

Top

Biography

Jong-Hwan Kim received his B.S., M.S., and Ph.D. degrees in Electronics Engineering from Seoul National University, Korea, in 1981, 1983 and 1987, respectively.

Since 1988, he has been with the Department of Electrical Engineering and Computer Science at the Korea Advanced Institute of Science and Technology (KAIST), where he is currently a Professor. His current research interests are in the areas of ubiquitous and genetic robotics.

He currently serves as an Associate Editor of the IEEE Transactions on Evolutionary Computation and of the IEEE Computational Intelligence Magazine. He was one of the co-founders of the International Conference on Simulated Evolution and Learning (SEAL). He was General Chair for the IEEE Congress on Evolutionary Computation in Seoul, Korea, 2001.

His name was included in the Barons 500 Leaders for the New Century in 2000 as the Father of Robot Football. He is the Founder of FIRA (The Federation of International Robosoccer Association, www.FIRA.net) and IROC (The International Robot Olympiad Committee, www.IROC.org). He is currently serving FIRA and IROC as President.

Top

Fred Glover

What is evolution's most effective contribution to solving difficult problems? A partial answer might be: the human brain and its special mechanisms for exploiting adaptive memory.

From this perspective, striking successes in Evolutionary Computation are occurring by "hybrid" approaches that incorporate adaptive memory principles, especially those from Tabu Search (also sometimes called Adaptive Memory Programming). New records are being established in many areas by such designs.

We survey applications where these successes are occurring, ranging from bioinformatics to combinatorial optimization. We additionally demonstrate that a small number of strategies devoted to exploiting adaptive memory account for a significant portion of these successes.

Noteworthy advances, as we show, are now emerging in applications involving risk and uncertainty, as well as in more classical areas of discrete and non-linear optimization, and commercial software incorporating these advances is currently used by more than 60,000 different licensed users in engineering and industry.

Biographies

Peng-Yeng Yin received his B.S., M.S. and Ph.D. degrees in Computer Science from National Chiao Tung University, Hsinchu, Taiwan. From 1993 to 1994, he was a visiting researcher at the Department of Electrical Engineering, University of Maryland, College Park (UMD). In 2000, he was a visiting Professor at the Department of Electrical Engineering, University of California, Riverside (UCR). From 2006 to 2007, he was a visiting Professor at Leeds School of Business, University of Colorado. From 2001 to 2003, he was a Professor at the Department of Computer Science and Information Engineering, Ming Chuan University, Taoyuan, Taiwan. Since 2003, he has been a Professor of the Department of Information Management, National Chi Nan University, Nantou, Taiwan. Dr. Yin is a member of the Phi Tau Phi Scholastic Honor Society. He has published dozens of articles in reputable journals including IEEE Trans. on Pattern Analysis and Machine Intelligence, IEEE Trans. on Education, Pattern Recognition, Annals of Operations Research, etc. His current research interests include artificial intelligence, pattern recognition, content-based image retrieval, relevance feedback, machine learning, and computational intelligence.

Fred Glover is the MediaOne Chaired Professor of Systems Science and a Distinguished Professor of the University of Colorado system.  His research interests are in the applications of computer decision support systems, including industrial planning, financial analysis, systems design, energy and natural resources planning, logistics, transportation and large-scale allocation models. His work is embodied in computer software systems currently serving more than 60,000 users in the United States and many others abroad. He has authored or co-authored more than 360 published articles and eight books in the fields of mathematical optimization, computer science and artificial intelligence. He is also the originator of Tabu Search (Adaptive Memory Programming), an optimization search methodology on which more than 500,000 Web pages can be found with a simple Google search.

Professor Glover is the recipient of the highest honor of the Institute of Operations Research and Management Science, the von Neumann Theory Prize, and serves as an elected member of the National Academy of Engineering. He has also received numerous other awards and honorary fellowships, including those from the American Association for the Advancement of Science (AAAS), the NATO Division of Scientific Affairs, the Institute of Operations Research and Management Science (INFORMS), the Decision Sciences Institute (DSI), the U.S. Defense Communications Agency (DCA), the Energy Research Institute (ERI), the American Assembly of Collegiate Schools of Business (AACSB), Alpha Iota Delta, the Institute of Cybernetics of the Ukrainian Academy of Science and the Miller Institute for Basic Research in Science. He also serves on the advisory boards of several organizations and is co-founder of OptTek Systems, Inc.

Top

In the recent years, numerous algorithms taking inspiration from nature have been proposed to handle continuous optimization problems: real-coded Genetic Algorithm using some specific variation operators (e.g. PCX crossover), Evolution Strategies using Gaussian mutations with adaptive or self-adaptive update strategies, Differential Evolution, and Particle Swarm Optimization, to name a few.

On the other hand, numerical algorithms have been proposed and improved for decades by applied mathematicians, relying on theoretical studies that guarantee their convergence under some (more or less) restrictive hypotheses. Those methods are readily available in all numerical toolboxes, and bio-inspired methods cannot hope to become widely accepted by engineers and scientists before demonstrating some added-value compared to those more conventional methods.

As one can expect, the choice of the best optimizer is problem-dependent. Nevertheless, some characteristics of the problem at hand have been identified for their impact on the performances of optimization methods: Multi-modality (existence of several local optima), often the only justification for the use of stochastic methods; Separability among the variables (problem can be solved one variable at a time); Ill-conditioning (slopes of the fitness landscape in different directions varies by several orders of magnitude); And noise in the objective function. Those characteristics are present in many real-world problems, but are difficult to a priori measure on a given black-box objective function.

This talk will survey some of the most popular bio-inspired and mathematical optimization methods for continuous problems. A priori analyzes and experimental comparisons will be presented to study their behavior and assess their robustness when facing multi-modal, ill-conditioned, non-separable and/or noisy functions.

Biography

Marc Schoenauer is Senior Researcher (Directeur de Recherche) with INRIA, the French National Institute for Research in Computer Science and Control. He graduated at Ecole Normale Supèrieure in Paris, and obtained a PhD in Numerical Analysis at Université Paris 6 in 1980. From 1980 until Aug. 2001 he has been full time researcher with CNRS (the French National Research Center), working at CMAP (the Applied Maths Laboratory) at École Polytechnique. He then became Directeur de Recherche at INRIA and worked in the Projet Fractales, before founding the TAO team in September 2003 together with Michèle Sebag.

Marc Schoenauer has been working in the field of Evolutionary Computation since the early 90s, is author of more than 60 papers in journals and major conferences of that field. He is or has been advisor of 18 PhD students.
He has also been part-time Associate Professor at Ecole Polytechnique in the Applied Maths Department from 1990 to 2004, and was in charge of the Optimization track at Ecole Nationale des Ponts et Chaussées from 2001 to 2005 and is currently teaching the Master2 course Evolutionary Computation and Robotics at Université Paris-Sud.

Marc Schoenauer is Senior Fellow and member of the Board of the ISGEC (International Society of Genetic and Evolutionary Computation), that has now become SIGEVO, the ACM Special Interest Group for Evolutionary Computation. He was member of the Executive Committee of the European Network of Excellence on Evolutionary Computation (Evonet) since its first funding in 1996 until its end in 2003, has served in the IEEE Technical Committee on Evolutionary Computation from 1995 to 1999, and is member of the PPSN Steering Committee. He was the founding president (1995-2002) of Evolution Artificielle, the French Society for Evolutionary Computation, who organizes the series of conferences Evolution Artificielle. He has been member of numerous program committees of international conferences, and was general chair of PPSN'2000 conference in Paris.

Marc Schoenauer is Editor in Chief of Evolutionary Computation Journal, has been associate editor of the IEEE Transactions on Evolutionary Computation from its start in 1996 until 2004, and of the Theoretical Computer Science - Theory of Natural Computing (TCS-C) journal since its creation in 2001 until 2006. He is still associate editor of the Genetic Programming and Evolvable Machines Journal, of the Journal of Applied Soft Computing.

Top

Evolutionary multiobjective optimization (EMO) is one of the most active research areas in the field of evolutionary computation. Various EMO algorithms have been proposed and successfully applied to a number of real-world problems. Recently developed well-known and frequently-used EMO algorithms such as NSGA-II and SPEA2 can be characterized by the use of three ideas: Pareto dominance-based fitness evaluation, diversity maintenance, and elitism. Whereas these EMO algorithms work very well on multiobjective optimization problems with a few objectives, their search ability severely deteriorates when the number of objectives increases. That is, usually Pareto-based EMO algorithms do not work well on many-objective optimization problems. Almost all hot issues in the design of EMO algorithms are related to the handling of many-objective optimization problems.

In this talk, we first explain the difficulty in the handling of many-objective optimization problems by Pareto dominance-based EMO algorithms. Next we show some approaches for improving the scalability of EMO algorithms to many-objective optimization problems such as the modification of Pareto dominance relation, the use of a performance indicator of non-dominated solution sets (e.g., hypervolume measure), the incorporation of the decision maker’s preference into EMO algorithms, the use of scalarizing functions, and the hybridization with local search. Then we show some promising application areas of EMO algorithms such as multiobjective machine learning and multiobjective scheduling.

Biography

Hisao Ishibuchi received his B.S. and M.S. degrees from Kyoto University in 1985 and 1987, respectively. He received the Ph. D. degree from Osaka Prefecture University in 1992. Since 1987, he has been with Osaka Prefecture University where he is currently a professor at Department of Computer Science and Intelligent Systems. He is also the head of Computational Intelligence Research Center of Osaka Prefecture University. His research interest includes evolutionary multiobjective optimization, multiobjective memetic algorithms, evolutionary scheduling, evolutionary game, genetic fuzzy systems, and fuzzy rule-based classification systems.

Dr. Ishibuchi is currently an associate editor of the IEEE Transactions on Evolutionary Computation, the IEEE Transactions on Fuzzy Systems, the IEEE Transactions on Systems, Man and Cybernetics - Part B, and the IEEE Computational Intelligence Magazine. He is also a Vice Chair of the IEEE CIS Fuzzy Technical Committee. He was an Area Chair of IJCNN 1997 and FUZZ-IEEE 1998, a Program Co-Chair of FUZZ-IEEE 2006 and EMO 2007, and will serve as the Program Chair for CEC 2010.

Top

Anthropologists have long recognized the importance of Culture as a symbolic entity that emerges from individuals experiences, co-evolves with the population, and in turn influences individual choices. How and why different Cultural forms have evolved in some environments and not others have always attracted the interest of scholars. In this talk Cultural Algorithms are employed a modeling framework in which to address the emergence of complex cultural structures and to study their impact on a population of individuals. It addresses both theoretical and practical issues in the development of large-scale multi-agent systems that contain large amounts of Cultural knowledge.

Biography

Robert G. Reynolds (M ’80) received his masters and Ph.D. degrees in Computer Science from the University of Michigan, Ann Arbor, in 1978 and 1979 respectively. He is currently a professor of Computer Science at Wayne State University, Detroit, MI, and an Adjunct Associate Research Scientist in the Museum of Anthropology at the University of Michigan, Ann Arbor. Dr. Reynolds heads the Artificial Intelligence Laboratory in the department of Computer Science at Wayne State. His NSF-funded research interests include computational models of the evolution of complex multi-agent societies, evolutionary learning, genetic algorithms, and complex systems. He has developed a class of computational models that describe the cultural evolution process, cultural algorithms. Dr. Reynolds is the co-author of three books on cultural system modeling: Acquisition of Software Engineering Knowledge: Sweep, An Automatic Programming System Based Upon Genetic Programming and Cultural Algorithms with George Cowan, Flocks of the Wamani with Joyce Marcus and Kent Flannery, and most recently Excavations at San Jose Mogote 1: The Household Archaeology with Kent Flannery and Joyce Marcus. He has also published articles in IEEE Software, Communications of the ACM, IEEE Transactions on Evolutionary Computation, IEEE Transactions on Systems, Man, and Cybernetics, Proceedings of the National Academy of Science, and Scientific American among others.

Top

An artificial brain is defined to be a network of evolved neural network modules, with each module performing its own evolved little job. A neural net module is evolved in a Celoxica electronic accelerator board 50 times faster than in an ordinary PC. Such a board costs only about $1500. Each evolved module is downloaded from the board to the memory of the PC. This is done 10,000s of times, with each module having its own fitness definition. The modules are then interconnected in the PC (e.g. the output signal of module M2369 becomes the second input signal of module M5620) according to the designs of human BAs (“Brain Architects”) to form artificial brains. Today’s PCs can perform the neural signaling of roughly 50,000 modules sequentially in real time (i.e. 25 signals per artificial neuron). The artificial brain (in the PC) is then used to control the many behaviors of an intelligent robot, via 2-way radio antenna. This brain building approach is both quick and inexpensive, so we hope other research groups will adopt it.

Biography

Hugo de Garis is a full professor of Computer Science, Pure Mathematics, and Theoretical Physics, in the International School of Software, Wuhan University, Wuhan, Hubei Province, China. Previously he  was an associate professor of Computer Science and adjunct associate professor of mathematical physics at Utah State University, in Utah, USA. He is a bit of a cultural adventurer, having lived in 7 countries, and chooses to live in China now and experience its rapid rise, before he gets too old. His research interests are in artificial brains, and his teaching is aimed at converting computer science graduate students into mathematical physicists, so that they can cope with the "topological quantum computing revolution" that demands PhD level math and  physics knowledge. Prof de Garis is the  author of "The Artilect War : Cosmists vs. Terrans : A Bitter Controversy Concerning Whether Humanity Should  Build Godlike Massively Intelligent Machines", published in the US and  China. Prof de Garis has recently appeared in TV documentaries (on the rise of massive artificial intelligence) in the US (ABC) and the UK (BBC).

Top