Abstracts and bios for the tutorials are here. The tutorials will all be on December 9.
| Main |
| Papers and Past |
| Program |
| Tutorials |
| Invited Talks |
| Workshops |
| NEW |
| NEW |
| Invitation Letters |
| Volunteering |
9:30-11:30am
Recently there has been renewed interest in game theory in several research disciplines, with its uses ranging from the modeling of evolution to the design of distributed protocols. In the AI community, game theory is emerging as the dominant formalism for studying strategic and cooperative interaction in multi-agent systems.
Classical work provides rich mathematical foundations and equilibrium concepts, but relatively little in the way of computational and representational insights that would allow game theory to scale up to large, complex systems. The rapidly emerging field of computational game theory is addressing such algorithmic issues, and this tutorial will provide a survey of developments so far. As the NIPS community is well-poised to make significant contributions to this area, special emphasis will be placed on connections to more familiar topics. Topics to be covered include:
- Basics of classical game theory: zero- and general-sum matrix games, Nash equilibria, minimax and fixed point arguments, linear programming
- Alternative equilibrium concepts: correlated equilibria, cooperative games and bargaining
- Evolutionary game theory
- Compact graphical representations of multi-player games, with connections to probabilistic inference in graphical models
- Multi-stage and stochastic games, with connections to reinforcement learning
- Learning in games: fictitious play, gradient algorithms, exponential update methods, with connections to machine learning
- Sampling arguments in game theory
- Game theory and distributed algorithms
Michael Kearns
is a professor in the Computer and Information Science
department at the University of Pennsylvania, and co-director of
Penn's Institute for Research in Cognitive Science. His research
interests include probabilistic models and algorithms in artificial
intelligence, computational game theory, computational learning
theory, and related topics. Prior to joining the Penn faculty, Kearns
spent a decade in basic research at AT&T/Bell Labs, where he headed
the AI and machine learning research department. He has published
widely in AI and ML, and is co-author of "An Introduction to
Computational Learning Theory" (with U. Vazirani, MIT Press 1994).
9:30-11:30am
Newton and Leibniz taught us how to integrate, and then argued over who should get the credit. Little did they know they had been scooped. Even when unschooled in calculus, the brains of humans and other animals naturally carry out the mathematical operation of integration with respect to time. This capability evolved because temporal integration is important for certain behaviors, such as motor control and navigation. Researchers have identified particular brain areas that function as neural integrators. A physiological explanation of how neurons integrate is still lacking, although a number of hypotheses have been proposed. Integration can be regarded as the simplest form of working memory, the ability to store information and actively manipulate it to perform computations. Therefore, understanding how neurons integrate could shed light on how working memory is implemented by the brain.
Sebastian Seung
is Assistant Investigator of the Howard Hughes Medical
Institute and Robert A. Swanson Career Development Professor in Life
Sciences in the Department of Brain and Cognitive Sciences and the
Department of Physics at the Massachusetts Institute of Technology. He
studied theoretical physics with David Nelson at Harvard University
and completed postdoctoral training with Haim Sompolinsky at the
Hebrew University of Jerusalem. Before joining the MIT faculty, he was
a member of the Theoretical Physics Department at Bell
Laboratories. He is a Sloan Research Fellow, a Packard Fellow, and a
McKnight Scholar.
Serge Belongie, UC San Diego
Jianbo Shi, Carnegie Mellon University
1:00-3:00pm
Spectral methods are a general class of algorithms that approximate computationally hard problems via properties of the eigenspectrum. These methods have been used for graph partitioning for over 30 years and have more recently shown great promise in a range of applications including VLSI design and computer vision.
The tutorial will cover the theory and practice of eigenvector methods for clustering and image segmentation. We will discuss the general theory of spectral methods as well as particular issues that arise when dealing with images: e.g. calculating affinity matrices from pixels and efficiently computing eigenvectors in spaces with a million dimensions.
Jianbo Shi
was born in Shanghai, China. He received his B.A. in
Computer Science and Mathematics from Cornell in 1994 and his Ph.D. in
Computer Science from UC Berkeley in 1998. In 1999, he joined the
research faculty of the Robotics Institute at Carnegie Mellon
University, where he is currently the joint PI and head of the HumanID
project. Starting January 2003, he will be an assistant professor in
the Department of Computer and Information Science at the University of
Pennsylvania. His current research interests include topics in image
segmentation, object recognition, and human identification-activity
inference. He has also conducted research in motion tracking, 3D
reconstruction, image retrieval, and humanoid robotics.
Serge Belongie
is an assistant professor in the Department of Computer
Science and Engineering at U.C. San Diego. He received his B.S. in
Electrical Engineering from Caltech in 1995 and his Ph.D. in
Electrical Engineering and Computer Sciences (EECS) from U.C. Berkeley in
2000. He is a co-founder of Digital Persona, Inc., and the principal
architect of the Digital Persona fingerprint recognition algorithm. His
research interests include computer vision, pattern recognition, and
digital
signal processing.
Yair Weiss
is a senior lecturer at the Hebrew University School of
Computer Science and Engineering. He received his Ph.D. from the
Massachusetts Institute of Technology and was a visiting scientist at
U.C. Berkeley. His research interests include human and machine
vision, machine learning and error correcting codes.
1:00-3:00pm
A fundamental goal of biology is to understand life at the level of genes, proteins and cells. Molecular biology and genetics are undergoing revolutionary changes. Emphasis has shifted from the study of individual genes and proteins to the exploration of the entire genome of an organism and the study of networks of genes and proteins. As the level of aspiration rises and the amount of available data grows by orders of magnitude, the field becomes increasingly dependent on mathematical and statistical modeling, mathematical analysis and computation. We shall give an introduction to the mathematical and computational challenges that arise in this field, with an emphasis on discrete algorithms and the role of combinatorics, optimization, probability, statistics, pattern recognition and machine learning.
Specific topics to be covered will be drawn from the following list: alignment and comparison of biological sequences; sequence assembly algorithms; the use of Hidden Markov Models (HMMs) to describe sequence families; gene finding using HMMs, pair HMMs and whole-genome alignment; discovery of protein binding sites using the EM algorithm; phylogeny construction using parsimony methods, distance-based methods and maximum likelihood methods; combinatorics of genome rearrangement; unsupervised analysis of gene expression data using feature selection, clustering and two-dimensional clustering; supervised analysis of gene expression data using support vector machines.
No prior knowledge of molecular biology will be assumed.
Richard M. Karp
was born in Boston, Massachusetts on January 3, 1935. He
attended Boston Latin School and Harvard University, receiving his Ph.D.
in 1959. From 1959 to 1968, he was a member of the Mathematics
Department at the IBM Thomas J. Watson Research Center. From 1968 to
1994, he was a Professor at the University of California, Berkeley, where
he held the Class of 1939 Chair. From 1988 to 1995, he was a Research
Scientist at the International Computer Science Institute (ICSI) in
Berkeley. From 1995 to 1999, he was a Professor of Computer Science and
Adjunct Professor of Molecular Biotechnology at the University of
Washington. In 1999, he returned to ICSI and Berkeley, where he is a
University Professor with appointments in Computer Science, Mathematics
and Bioengineering. During the 1999-2000 academic year, he was the
Hewlett-Packard Visiting Professor at the Mathematical Sciences Research
Institute in Berkeley.The unifying theme in Karp's work has been the study of combinatorial algorithms. His 1972 paper "Reducibility Among Combinatorial Problems" showed that many of the most commonly studied combinatorial problems are NP-complete, and hence likely to be intractable. Much of his subsequent work has concerned parallel algorithms, the probabilistic analysis of combinatorial optimization algorithms and the construction of randomized algorithms for combinatorial problems. His current activities center around algorithmic methods in genomics and computer networking. He has supervised thirty-five Ph.D. dissertations.
Karp has received the U.S. National Medal of Science, the Turing Award, the Fulkerson Prize, the Harvey Prize (Technion), the Centennial Medal (Harvard), the Lanchester Prize, the Von Neumann Theory Prize, the Von Neumann Lectureship, the Distinguished Teaching Award (Berkeley), and the Babbage Prize. He holds six honorary degrees. He is a member of the U.S. National Academies of Sciences and Engineering and a Fellow of the American Academy of Arts and Sciences, the American Association for the Advancement of Science and the Association for Computing Machinery.
3:30-5:30pm
How do listeners recover information about individual sound sources from the acoustic melange reaching the ears? From musical ensembles to conversations in noisy settings, everyday listening situations present a significant challenge to our auditory systems and recognition algorithms alike. Cues such as pitch and location are prime candidates for source separation approaches, but prior expectations also have a role.
Armed with a variety of audio demonstrations, this tutorial will explore the perceptual organisation of sound from the perspective of auditory physiology, psychophysics and computational modelling. Topics to be covered include neurally-inspired models for the extraction of useful signal features, architectures for feature-binding and missing-data approaches to robust automatic speech recognition. The tutorial will also look in detail at the theory and practice of a recent probabilistic decoder which integrates bottom-up and top-down processes in computational auditory scene analysis.
After a spell at the UK National Physical Laboratory,
Martin Cooke
obtained his PhD in 1991 and is currently Reader in Computer Science
at the University of Sheffield. His interests lie in speech
perception, computational auditory scene analysis, robust automatic
speech recognition and educational tools for speech, hearing and
phonetics. He is the author of Modelling Auditory Processing and
Organisation (Cambridge) and co-edited Visual Representations of
Speech Signals (Wiley).
William Cohen, Carnegie Mellon University, CALD
3:30-5:30pm
The Web is the world's largest knowledge base. However, its data is in a form intended for human reading, not manipulation, data mining and reasoning by computers. Information extraction is the process of filling fields in a database by automatically extracting sub-sequences of human readable text. Today's search engines return web pages. Tomorrow's search engines will use information extraction to return "things" (like people, jobs, companies, events), their relations, facts and trends. This tutorial will survey many of the sub-problems and methods of information extraction, including use of sliding-window and finite state machines, language and formatting features, generative and conditional models, rule-learning and Bayesian techniques. We will also discuss some related issues, such as association of data fields into records, reference matching and de-duplication. A familiarlity with statistical learning techniques, such as maximum likelihood estimation, Bayesian networks, and hidden Markov models, would be useful, but is not required.
Andrew McCallum
is a Research Associate Professor at University of
Massachusetts, Amherst. He was previously Vice President of Research
and Development at WhizBang Labs, a company that used machine learning
for information extraction from the World Wide Web. In the late
1990s, he was a Research Coordinator at Justsystem Pittsburgh Research
Center. He received his PhD in computer science from University of
Rochester in 1995 and was a post-doctoral fellow at Carnegie Mellon
University in 1996. He is on the editorial board of the Journal of
Machine Learning Research and has co-organized numerous technical
workshops. For the past seven years, McCallum has been active in
research on statistical machine learning applied to text, especially
information extraction, document classification, finite state models,
and learning from combinations of labeled and unlabeled data.
William Cohen
received his bachelor's degree in Computer Science from
Duke University in 1984 and his PhD in Computer Science from Rutgers
University in 1990. From 1990 to 2000, Cohen worked at AT&T
Labs-Research, then later at Whizbang Labs, a company specializing in
information extraction from the World Wide Web. Cohen is currently an
action editor for the Journal of Machine Learning Research and
previously served as an editor for the journal Machine Learning and
the Journal of Artificial Intelligence Research. He co-organized the
1994 International Machine Learning Conference and has served on more
than 20 program or advisory committees. His research interests include
information integration and machine learning, particularly text
categorization and learning from large datasets.
About this Webpage
For issues regarding page design and content, contact Alexander Gray. For issues regarding forms, scripts and server operation, contact Guy Lebanon.