Fifteenth Annual Conference on. Neural Information Processing Systems. Tutorials. December 3, 2001 Hyatt Regency Vancouver, British Columbia, Canada

Save this PDF as:
 WORD  PNG  TXT  JPG

Size: px
Start display at page:

Download "Fifteenth Annual Conference on. Neural Information Processing Systems. Tutorials. December 3, 2001 Hyatt Regency Vancouver, British Columbia, Canada"

Transcription

1 NIPS 2001 Fifteenth Annual Conference on Neural Information Processing Systems Tutorials December 3, 2001 Hyatt Regency Vancouver, British Columbia, Canada Conference Sessions December 4-6, 2001 Hyatt Regency Vancouver, British Columbia, Canada Workshops December 6-8, 2001 Delta Whistler Resort Whistler, British Columbia, Canada

2 NIPS*2001 is sponsored by the Neural Information Processing Systems Foundation, Inc. Foundation Officers President Terrence Sejnowski The Salk Institute Vice President for Development Gary Blasdel Harvard Medical School Treasurer Bartlett W. Mel University of Southern California Secretary Michael Mozer University of Colorado, Boulder Board Members Leo Breiman University of California, Berkeley Jack Cowan University of Chicago Stephen Hanson Rutgers University Michael I. Jordan University of California, Berkeley Michael Kearns AT&T Laboratories Research Scott Kirkpatrick Hebrew University, Jerusalem Todd K. Leen Oregon Graduate Institute Richard Lippmann Massachusetts Institute of Technology John Moody Oregon Graduate Institute Sara A. Solla Northwestern University Medical School Gerald Tesauro IBM Watson Labs Dave Touretzky Carnegie Mellon University Emeritus Members T. L. Fine Cornell University Eve Marder Brandeis University Organizing Committee General Chair Thomas G. Dietterich Oregon State University Program Chair Sue Becker McMaster University Publications Chair Zoubin Ghahramani University College London Tutorials Chair Yoshua Bengio University of Montreal Workshop Co-Chairs Virginia de Sa University of California, San Francisco Barak Pearlmutter University of New Mexico Publicity Chair Richard Zemel University of Toronto Volunteers Director Sidney Fels University of British Columbia Government Liaison Gary Blasdel Harvard Medical School Contracts Stephen Hanson Rutgers University Scott Kirkpatrick Hebrew University, Jerusalem Gerry Tesauro IBM Watson Labs Program Committee Sue Becker (Program Chair) McMaster University Gert Cauwenberghs Johns Hopkins University Bill Freeman Mitsubishi Electric Research Lab Thomas Hofmann Brown University Dan Lee Bell Laboratories Lucent Technologies Sridhar Mahadevan University of Massachusetts, Amhurst Marina Meila University of Washington Klaus Mueller GMD First, Berlin Klaus Obermayer Technische Universität, Berlin Sam Roweis University College London John Shawe-Taylor Royal Holloway University of London Josh Tenenbaum Stanford University Volker Tresp Siemens, Munich Richard Zemel University of Toronto NIPS Foundation Office The Salk Institute for Biological Studies North Torrey Pines Road La Jolla, CA Administrative Manager Rosemary Miller The Salk Institute Legal Advisor Phil Sotel Pasadena, CA Accountant Gabriele Larmon University of Southern California Information about the Conference is available via the World Wide Web at: Co-Webmasters: Alexander Gray, Carnegie Mellon University Xin Wang, Oregon State University

3 NIPS*2001 Fifteenth Annual Conference on Neural Information Processing Systems You are invited to participate in the Fifteenth Annual Conference on Neural Information Processing Systems, which is the premier scientific meeting on neural computation. The Conference is extraordinarily interdisciplinary, with contributions from many intellectual communities united by an interest in the study of artificial and natural neural information processing. Interdisciplinary interactions are encouraged by having a single track of presentations and high-quality poster sessions. Presentation topics include learning algorithms and architectures, learning theory, neuroscience, cognitive science, vision, speech and signal processing, reinforcement learning and control, implementations, and diverse applications. All papers have been rigorously reviewed (with a 30% acceptance rate). The Conference is preceeded by one day of tutorials and is followed by two days of workshop sessions in nearby Whistler, British Columbia, Canada. The Tutorial Program provides a choice of eight two-hour tutorials by leading scientists. The topics span a wide range of subjects including neuroscience, learning theory, SVM and Bayesian learning methods, as well as applications to language processing and robotics. The Workshop Program includes 19 workshops covering a wide range of topics from neuroimaging and brain-computer interfaces to kernel-based learning methods and the 2001 Unlabeled Data for Supervised Learning Competition. The workshop schedule allows time for informal discussions, skiing, and other winter sports. Program Events Tutorials: Monday, December 3 8:00 am 6:00 pm Registration 9:30 am 5:30 pm Tutorials 7:00 pm 10:00 pm Conference Banquet Conference: Tuesday, December 4 & Wednesday, December 5 8:00 am 6:00 pm Registration 8:30 am 12:00 pm Sessions 2:00 pm 5:15 pm Sessions 10:00 am 6:00 pm Poster Setup and Preview 7:30 pm 10:30 pm Poster Session Thursday, December 6 8:30 am 12:00 pm Sessions Workshops: Thursday, December 6 2:00 pm 3:30 pm Buses Depart for Workshops 6:30 pm 8:30 pm Welcoming Reception and Registration at Delta Whistler Friday, December 7 7:00 am 11:00 am Registration 7:30 am 10:30 am Workshop Sessions 4:00 pm 7:00 pm Workshop Sessions Saturday, December 8 7:30 am 10:30 am Workshop Sessions 4:00 pm 7:00 pm Workshop Sessions 7:30 pm 10:30 pm Banquet & Wrap Up Meeting 1

4 Program Highlights Tutorials Monday, December 3, 2001 Session 1: 9:30 11:30 am Shawn Lockery, University of Oregon Why the Worm Turns: How to Analyze the Behavior of an Animal and Model Its Neuronal Basis Luc Devroye, McGill University Nonparametric Density Estimation: VC to the Rescue Session 2: 1:00 3:00 pm Bernhard Schölkopf, Biowulf Technologies and Max-Planck-Institute for Biological Cybernetics Support Vector Machines and Kernel Methods Daphne Koller, Stanford University and Nir Friedman, Hebrew University Learning Bayesian Networks From Data Session 3: 3:30 5:30 pm Christopher Manning, Stanford University Probabilistic Linguistics and Probabilistic Models of Natural Language Processing Tutorial Abstracts may be found on pages 4 and 5. Sebastian Thrun, Carnegie Mellon University Probabilistic Robotics Conference Sessions Tuesday, Wednesday, Thursday, December 4-6, 2001 Over 650 technical papers have been submitted for the 2001 Conference. Each was carefully reviewed by three referees. Twenty-five papers were accepted for oral presentation, and 171 papers were accepted for poster presentation. All accepted papers will appear in the Proceedings, which will be available online, on CD-ROM, and as a book from MIT Press. The CD-ROM version will be distributed to everyone who registers for the Conference. The MIT Press book must be ordered separately on the registration form on page 9. Invited Speakers Barbara Finlay, Cornell University How Brains Evolve and the Consequences for Computation Alison Gopnik, University of California, Berkeley Babies and Bayes-Nets: Causal Inference and Theory-formation in Children, Chimps, Scientists, and Computers Jon M. Kleinberg, Department of Computer Science, Cornell University Decentralized Network Algorithms: Small-World Phenomena and the Dynamics of Information Tom Knight, Artificial Intelligence Laboratory, MIT (Banquet Speaker) Computing With Life Judea Pearl, University of California, Los Angeles Causal Inference as an Exercise in Computational Learning Shihab A. Shamma, University of Maryland, College Park Common Principles in Auditory and Visual Processing 2

5 Workshop Sessions Friday and Saturday, December 7 and 8, 2001 NIPS Workshops provide multi-track intensive sessions on topics ranging from computational learning theory to experimental neuroscience. The venue and schedule facilitate interaction, informality, and depth. The workshops will be held at the Delta Whistler Resort and the Whistler Conference Center in Whistler, British Columbia, Canada. The workshop sessions will end with a banquet on Saturday evening, December 8, Scheduled Workshops Machine Learning/Bioinfomatics Innovative Machine Learning Methods for Information Access and Management Artificial Neural Networks in Safety-Related Areas: Applications and Methods for Validation and Certification Activity-Dependent Synaptic Plasticity Information and Statistical Structure in Spike Trains Neuroimaging: Tools, Methods and Modeling Directions in Brain-Computer Interface Research Quantum Neural Computing Geometric Methods in Learning Minimum Description Length New Directions in Kernel-Based Learning Methods Variable and Feature Selection Adequate Forms of Knowledge Representation in Meta- Learning The 2001 Unlabeled Data for Supervised Learning Competition Learning Methods in Multi-Sensory (Audio+Video) Perception New Methods for Preference Elicitation Computational Neuropsychology Causal Learning and Inference in Humans and Machines Foundations of Occam s Razor and Parsimony in Learning Additional information on scheduled workshops can be found on the NIPS*2001 web site. Instructions for Poster Presentation Posters will be on view from 7:30 to 10:30 pm on Tuesday, December 4, and Wednesday, December 5, You should set up your poster during the period from 10:00 am to 6:00 pm on the day of your poster presentation. Posters will be assigned numbers (e.g. AA34). These numbers will be attached to the individual display boards in the Poster Presentation Room. You should find the display board corresponding to your number and attach your poster there. Poster pins will be provided. Each poster presentation will occupy a 4 foot x 8 foot surface area. Plan your poster to fit this space. Identify your poster near the top of the panel with title and name(s) of investigator(s). All typefaces and graphics on the poster should be large enough to be read comfortably from distances of four to five feet. A well-designed poster clearly states the main problem and contribution and guides the viewer through the presentation. Hardcopies of viewgraphs do not make a good poster. A matte finish on your poster materials will give better visibility under the glare from lights. You may want to have copies of your paper or poster, along with other material to give to viewers. Tuesday night posters must be removed by 9:00 am on Wednesday morning. Wednesday night posters must be removed by 9:00 am on Thursday morning. 3

6 The complete description of the morphology and synaptic connectivity of all 302 neurons in the nematode worm C. elegans raised the prospect of the first comprehensive neurobiological account of the entire behavioral repertoire of an organism. Using C. elegans chemotaxis and thermotaxis as examples, this tutorial takes a realistic, quantitative look at the three-step process of neuro-behavioral analysis: (1) algorithm discovery, (2) systems identification, and (3) neuronal implementation. This process will be illustrated by real-life examples of automated behavioral tracking, laser ablation of neurons, linear systems analysis, patch-clamp electrophysiology, simulated annealing of neural network models, and robotics. The outlook for a similar analysis of the vertebrate nervous system will be addressed. Shawn Lockery is an Associate Professor in the Institute of Neuroscience and Department of Biology at the University of Oregon. He did his Tutorial Abstracts Session 1 9:30 11:30 am Shawn Lockery, University of Oregon Why the Worm Turns: How to Analyze the Behavior of An Animal and Model Its Neuronal Basis graduate work in philosophy and psychology at Oxford University and in neurobiology at the University of California, San Diego, where he worked with William Kristan on the neuronal basis of behavior in the leech. Lockery did his postdoctoral work in neural network modeling with Terrence Sejnowski at The Salk Institute. Lockery s present research investigates the neuronal basis of spatial orientation behaviors in the nematode C. elegans. His research program combines quantitative behavioral analysis, electrophysiology, modeling, and robotics to understand adaptive control in dynamical networks. Lockery is the recipient of the NSF New Young Investigator Award, Searle Scholars Program Award, ONR Young Investigator Award, and Biology Teacher Recognition Award, University of Oregon. Web page: Luc Devroye, McGill University Nonparametric Density Estimation: VC to the Rescue Estimating a distribution function is easy, but estimating a density is not. A density can only be estimated well if the data are properly smoothed, a process that is typically controlled by a few (smoothing) parameters. Most research on density estimation relates the choice of these parameters to performance. In the tutorial, we suggest a universal methodology for automatic (data-based) smoothing. The NIPS*2001 crowd will be happy to learn that the performance of the density estimates is related to the VC dimension of certain classes of sets. In fact, for many estimates, the method is nearly optimal in the following sense: For any density, no matter how pathological, the expected total variation error is asymptotically at most three times the optimal error (the error we would obtain if we knew the density beforehand but were forced to use the given density estimate). No other automatic smoothing methods we know of have this property. Luc Devroye is Professor of Computer Science at McGill University in Montreal. He has contributed in the areas of nonparametric estimation, random number generation, and probabilistic analysis of algorithms. He has written six books, including A Probabilistic Theory of Pattern Recognition (Springer-Verlag, 1996, with Gyorfi and Lugosi). The tutorial is based on his most recent book, Combinatorial Methods in Density Estimation (Springer-Verlag, 2001). Web page: cgm.cs.mcgill.ca/~luc/ In the 1990s, a new type of learning algorithm was developed, based on results from statistical learning theory: the Support Vector Machine (SVM). This gave rise to the development of a new class of theoretically elegant learning machines which use a central concept of SVMs kernels for a number of different learning tasks. Kernel machines now provide a modular and simple to use framework that can be adapted to different tasks and domains by the choice of the kernel function and the base algorithm, and they have been shown to perform very well in problems ranging from computer vision to text categorization and applications in computational biology. The tutorial will introduce the basic concepts necessary to appreciate the kernel approach, which is now becoming a standard part of the machine learning toolkit. Daphne Koller, Stanford University and Nir Friedman, Hebrew University Learning Bayesian Networks from Data Bayesian networks, or directed graphical models, are a compact and computationally efficient representation of multivariate probability distributions. In recent years, there has been significant progress in Session 2 1:00 3:00 pm Bernhard Schölkopf, Biowulf Technologies and Max-Planck-Institute for Biological Cybernetics SVM and Kernel Methods Bernhard Schölkopf is a Researcher at Biowulf Technologies (New York) and Director at the Max-Planck-Institute for Biological Cybernetics (Tübingen). Previously, he worked at AT&T Bell Labs, at GMD First, Berlin, at the Australian National University, at Microsoft Research Cambridge (UK), and taught at the Humboldt University and the Technical University Berlin. His scientific interests include machine learning and perception. He co-authored several books on SVMs and kernel methods, among them a textbook to be published in time for the Conference. He was awarded the Lionel Cooper Memorial Prize in Mathematics from the University of London, and the annual dissertation prize of the German Association for Computer Science (GI). He has recently been elected scientific member of the Max-Planck-Society. Web page: methods and algorithms for inducing Bayesian networks from data. Learning these particular models is desirable for several reasons. In particular, the learned models can be used directly for a variety of tasks, 4

7 including prediction, decision making, and diagnosis. The learned Bayesian network structure can also provide substantial insight into the dependency structure of the variables in the domain. In the first part of the tutorial, we will review the fundamental theory and algorithms for learning these networks from data, including parameter estimation, learning the dependency structure (model selection), and the treatment of missing data and hidden variables using the EM algorithm. In the second part of the tutorial, we will focus on advanced data analysis using Bayesian networks, including learning structure in the presence of hidden variables, learning structure from sparse data, and extensions to richer, heterogeneous data sets. We will describe applications of these techniques to a variety of data sets, including web data and genomic data. Prerequisite knowledge: Although the tutorial will be self-contained, we will review the basic concepts of directed graphical models only briefly. Familiarity with this representation, and with basic concepts in density estimation, is advised. Daphne Koller received her PhD from Stanford University in After two years of postdoctoral study at Berkeley, she returned to Stanford, where she is now an Associate Professor in the Computer Science Department. Her main research interest focuses on probabilistic models for complex systems, but covers representation, reasoning, and learning. She was the Co-Chair of the recent UAI 2001 Conference, has served on numerous program committees and editorial boards, and is currently an Associate Editor of the Journal of Artificial Intelligence Research. She was awarded the Sloan Foundation Faculty Fellowship in 1996, the ONR Young Investigator Award in 1998, the Presidential Early Career Award for Scientists and Engineers (PECASE) in 1999, and the IJCAI Computers and Thought Award at the recent IJCAI 2001 Conference. Web page: robotics.stanford.edu/~koller Nir Friedman received his PhD in Computer Science from Stanford in 1997, was a postdoctoral scholar in the Computer Science Division at the University of California, Berkeley until late 1998, and is currently a faculty member in the School of Computer Science and Engineering at the Hebrew University, Jerusalem. He has worked on inference and learning with probabilistic graphical models. In recent years, his work has focused on applications of these tools to computational biology. He is the Co-Chair of the upcoming UAI 2002 Conference, and has served on numerous program committees and editorial boards. He is also the recipient of Best Paper Awards at the recent UAI and ISMB (Intelligent Systems in Molecular Biology) Conferences. Web page: Session 3 3:30 5:30 pm Christopher Manning, Stanford University Probabilistic Linguistics and Probabilistic Models of Natural Language Processing This tutorial will cover (very briefly!): what the aims and goals of computational linguistics are; why there is a need for statistical and quantitative techniques in linguistics and computational linguistics, including connections between psycholinguistic work and computational models; the distinctive features of natural language from the perspective of statistical models; what the major practical problems in natural language understanding and generation are; and an overview of the methods that have been used to tackle these problems. A reasonable background in statistical learning, but less knowledge of NLP problems and standard ways of approaching them is assumed. The tutorial will focus on some particular areas of current interest, such as parsing, grammar induction, generative versus discriminative models and semantic representation. Christopher Manning is an Assistant Professor of Computer Science and Linguistics at Stanford University. He received his PhD from Stanford University in 1995, and served on the faculty of the Computational Linguistics Program at Carnegie Mellon University ( ), and the University of Sydney Linguistics Department ( ), before returning to Stanford. His research interests include probabilistic models of language, natural language parsing, constraint-based linguistic theories, syntactic typology, information extraction and text mining, and computational lexicography. He is the author of three books, including Foundations of Statistical Natural Language Processing (MIT Press, 1999, with Hinrich Schuetze). Web page: www-nlp.stanford.edu/~manning/ Sebastian Thrun, Carnegie Mellon University Probabilistic Robotics In recent years, Bayesian techniques have revolutionized the field of robotics. Many state-of-the-art robot systems rely on statistical techniques for coping with uncertainty, which arises naturally from sensor noise and environment dynamics. Examples of robotics problems that have recently been solved with Bayesian techniques include mobile robot localization, mapping, path planning, and navigation. This tutorial will provide a systematic introduction to the field of probabilistic robotics. Its purpose is to familiarize NIPS researchers with basic problems in robotics (solved and unsolved) and discuss principled statistical solutions. To the NIPS researcher, robotics offers a multitude of fascinating realworld problems, many of which involve high-dimensional parameter spaces and require advanced model selection techniques. I hope that this tutorial will enable participants to work on some of these problems, thereby increasing the impact of statistical techniques in an application domain that is fun and has the potential for high societal impact. Sebastian Thrun is an Associate Professor of Computer Science and Robotics at Carnegie Mellon University, and an Adjunct Associate Professor in the School of Nursing at the University of Pittsburgh. Thrun s research interests include machine learning and statistical inference, with a strong focus on advanced service robotics. Presently, Thrun is on sabbatical leave at Stanford University, where he is finishing an introductory book on the tutorial s topic. Web page: 5

8 Hotels Please note that all hotel rates are listed in Canadian dollars. The approximate rate in US dollars is indicated in parentheses. The conversion rate at the time of the preparation of this brochure: Local taxes of 17% will be added to all quoted hotel rates. When making your hotel reservations, be sure to mention that you are a member of the NIPS*2001 participant to receive the special group rate. Hyatt Regency Vancouver The Conference and Tutorials will be held at the Hyatt Regency Vancouver. The hotel is located in downtown Vancouver, overlooking a mountain-rimmed harbor. For accomodations at the Hyatt Regency contact: Hyatt Regency Vancouver Reservation Line Single/Double/Triple... $129 (~$83 US) 655 Burrard Street Reservation Fax Quadruple Occupancy... $156 (~$100 US) Vancouver, British Columbia Hotel Main Number Daily Parking Fee... $18.50 (~$12 US) Canada V6C 2R7 There will be an additional charge of $27 (~$17 US) per night for each additional person. Maximum room occupancy is four persons. A credit card or one night s room and tax pre-payment is needed to guarantee room reservations. The hotel has a 24-hour cancellation policy. An early departure fee of $75 (~$48 US) will be assessed if hotel is given less than 24-hour notice. Rooms will be held at the special NIPS*2001 rate until November 3, Beyond this date, the special rate is subject to space availability. Whistler Hotels Accommodations are available at the Delta Whistler Resort and the Crystal Lodge. Both are located just a short walking distance from each other and from the Whistler Conference Center. Workshops will be held at both the Delta Whistler Resort and the Whistler Conference Center. Attendees are strongly encouraged to make their hotel reservations before the specified cutoff date for each hotel, as rooms are in high demand during this period and likely to be sold out afterwards. Delta Whistler Resort The Delta Whistler Resort is located just next to the Conference Center and the Crystal Lodge. For accommodations at the Delta Whistler Resort contact: Delta Whistler Resort Reservation Line: Single/Double... $149 (~$96 US) 4050 Whistler Way Central Reservation: Deluxe with Kitchenette... $179 (~$115 US) Whistler, Reservation Fax: Junior Suite... $389 (~$250 US) British Columbia, Main Hotel Number: Daily Parking Fee... $14 (~$9 US) Canada V0N 1B Web site:... Rates at the Delta Whistler Resort are per night for single or double occupancy. There will be an additional charge of $30 (~$19 US) per night for each additional person. Rooms will be held for NIPS*2001 participants at the special rate until November 6, After that date, reservations will be accepted on a space-availability basis only. Full payment is required seven days in advance at the Delta Whistler Resort. There is no refund for early checkouts, late check-ins, or no-shows. Cancellations within 14 days of arrival will be charged for the full amount of the original reservation. Crystal Lodge The Crystal Lodge is located just two-minutes walk from the Delta Whistler Resort and the Conference Center. For accommodations at the Crystal Lodge contact: Crystal Lodge Reservation Line: Single/Double Standard Two Persons... $109 (~$71 US) 4154 Village Green Main Hotel Number: Single/Double Studio Two Persons... $136 (~$88 US) Whistler, Hotel Fax: Loft Four Persons... $154 (~$100 US) British Columbia, ... Three-Bedroom Suite Six Persons... $405 (~$263 US) Canada, V0N 1B4 Web site:... Daily Parking Fee... $8 (~$5 US) Rates are per night for single or double occupancy. There will be a charge of $15 (~$10 US) per night for each additional person. The Crystal Lodge requires full payment at the time of reservation. There is no refund for early checkouts, late check-ins, or no-shows. Cancellations received less than 30 days in advance of reservation cannot be refunded. After that, reservations will be based on availability and rates at that time; cancellation policies will apply. Rooms will be held for NIPS participants at the special group rate until November 9,

9 Discount Lift Tickets, Ski Rentals and Lessons Whistler: Lift tickets for the NIPS*2001 Workshops will be sold at a special group rate of $44 (~$28 US) (plus 7% GST per adult day ticket). Rentals, retail items, sightseeing tickets, and lessons will also be available at discounts ranging from 5-10%. The special discount rates are available for NIPS*2001 group members (including guests and family) and are only available when purchasing tickets at the Conference registration desk at the Delta Whistler Resort, during the opening night reception on Thursday, December 6 and during Friday morning registration hours. Local Transportation Options Vancouver International Airport is 13 miles from downtown Vancouver and the trip takes approximately 30 minutes, depending on traffic. It is possible to find a variety of means of transportation to downtown Vancouver just outside of both the domestic and international terminals. There is an AIRPORTER bus that leaves from these locations every 15 minutes. The cost for bus service is $12 Canadian (~$8 US) each way. You may phone for information or advance reservations at You may also take a taxi or limo for approximately $20-$25 Canadian (~$13-16 US). Rental cars and vans are available from car rental agencies that have offices at the airport. Note: Since most means of transportation will accept either US or Canadian dollars or credit cards, participants should be prepared to offer one of these methods of payment for local transportation. Driving Instructions: Vancouver International Airport to Hyatt Regency: (Driving time: ~25-40 minutes) Take Grant McConnachie Way over the Arthur Laing Bridge. Take the Granville Street Exit from bridge and continue straight down Granville Street (approximately 60 blocks). Turn left onto West 16th Avenue then turn right on Burrard Street. Continue down Burrard Street and over the Burrard Street Bridge. Continue on Burrard for approximately 8 blocks. You will see the Hyatt Regency Vancouver on the left. Pull into the front of the hotel and have your car valet parked or self-park. Hyatt Regency Vancouver to Whistler: (Driving time: ~2-2.5 hours) From the Hyatt Regency on Burrard, head South on Burrard and turn right (West) on Georgia and follow it through Stanley Park and over the Lions Gate Bridge. Exit west off/under the bridge and into West Vancouver. Turn right on Taylor Way at Park Royal Shopping Centre. Follow Taylor Way half-a-mile up the hill to the Highway 1 (Trans-Canada) overpass. Join Highway 1 Westbound until you reach the junction with Highway 99 (Sea-to-Sky Highway). Exit right onto Highway 99. Follow this route for just over 100 kilometers along scenic Howe Sound, past Squamish to Whistler. International Airport to Whistler: (Driving time: ~2.5 hours) Take Grant McConnachie Way (the main route out of the Vancouver International Airport) and over the Arthur Laing Bridge. Continue straight over the bridge, stay in the right hand lane, which connects to SW Marine Drive/Granville Street. Granville Street is a North- South connector leading to the heart of downtown. Once across the Granville Street Bridge, turn left onto Drake at the first set of lights. Turn right on Hornby Street and continue on Hornby until you reach Georgia Street. Turn left (West) on Georgia and follow it through Stanley Park and over the Lions Gate Bridge. Exit West off/under the bridge and into West Vancouver. Turn right on Taylor Way at Park Royal Shopping Centre. Follow Taylor Way half-a-mile up the hill to Highway 1 (Trans-Canada) overpass. Join Highway 1 Westbound until you reach the junction with Highway 99 (Sea-to-Sky Highway). Exit right on to Highway 99. Follow this route for just over 100 kilometers along scenic Howe Sound, past Squamish to Whistler. Bus from Hyatt Regency Vancouver to Whistler: Buses will be available to transport Conference participants from the Hyatt Regency Vancover to Whistler. These buses will leave from the Hyatt Regency on Thursday, December 6. Tickets may be purchased in advance on the registration form on page 9 or at the Conference registration desk. Return Transportation from Whistler: Each participant must make individual arrangements for the return from Whistler to the Vancouver International Airport. Perimeter's Whistler Express picks up at the Whistler Resort every 90 minutes beginning at 5:00 am and ending at 6:00 pm. The rate for NIPS*2001 participants is $46 (~$29 US). Reservations are required for this return service. Remember to identify yourself as a participant in the NIPS*2001 Conference to receive the special group rate. Make reservations no later than 48 hours prior to departure time to reserve your space. Phones: from US and Canada (Toll Free) from Whistler from Vancouver Reservations: 7

10 Registration Registration for the Main Conference includes a Monday evening welcoming banquet, continental breakfasts on Tuesday, Wednesday, and Thursday morning, as well as a CD-ROM containing the NIPS*2001 Conference Proceedings. Workshop registration includes opening night reception, Saturday evening banquet, and two full breakfasts. Registration for tutorials and workshops is optional and requires payment of a separate fee. A continental breakfast will be provided for tutorial participants. To register for NIPS*2001, please complete the registration form on the facing page and mail with full payment, or FAX to: NIPS*2001 Registration Fax: NIPS Foundation Office c/o The Salk Institute - CNL N. Torrey Pines Road La Jolla, CA USA Alternatively we offer on-line registration via the NIPS web site: Payments will be accepted in the form of check or money order made payable to the Neural Information Processing Systems Conference or by credit card (VISA or Mastercard). Payments must be in US dollars. When registering on-line, credit card payments may be made via a secure link. Registration will not be accepted without payment. Students will be asked to provide valid student identification at the NIP*2001 registration desk. WireTransfers: To be registered for the Conference we must receive your wire transfer to the Wells Fargo Bank no later than November 16, Please direct wire transfers to: Wells Fargo Bank International Wire Transfers: Wire Transfers within the USA: P. O. Box Swift Address: WFBIUS6S Routing Number: Sacramento, CA USA Neural Information Processing Systems Conference Account Number: Cancellation Policy: Cancellations must be received in writing. Cancellations received by November 2 will receive a full refund. Cancellations between November 3 and November 26 (inclusive), will be subject to a $75 (US) cancellation fee. No refund will be given for cancellations after November 26. Financial Support Very limited funding will be available to support the travel of young investigators, post-doctoral fellows, and graduate students to attend NIPS*2001. Awards will be based on merit, and heavily weighted by need. Travel awards will be in the amount of $400 (US) for travel from the US and within Canada and $600 (US) for participants arriving from elsewhere. Conference registration is not covered by travel awards. Applications for travel grants must be submitted via the NIPS Travel Grants form accessible through the Conference web site. Deadline for submission is midnight, Friday, October 12, Late applications will not be accepted. Award recipients will be notified by in late October. Travel award checks in US dollars will be available at the NIPS*2001 Conference registration desk. A copy of your airline ticket will be required to receive your travel award. Volunteers: Graduate students interested in doing volunteer work at the meeting in exchange for a registration fee waiver should consult the Volunteers entry on the Conference web site. Only a limited number of openings are available. Conference Proceedings The NIPS*2001 Conference Proceedings, Advances in Neural Information Processing Systems 14, edited by Thomas G. Dietterich, Sue Becker, and Zoubin Ghahramani, will be available to Conference attendees in three different formats. First, the complete Proceedings will be available on-line through the NIPS web site. Second, a CD-ROM version of the Proceedings will be mailed to each conference attendee in late Spring, Third, at their OPTION, attendees may purchase the Proceedings in a two-volume soft-cover format for the special Conference rate of $30 (US) by checking the appropriate box on the registration form. This special reduced rate is available only if the purchase is made before or during the Conference. After the Conference, copies of the hardcopy Proceedings may be obtained through: MIT Press Five Cambridge Center MIT Press orderline at Cambridge, MA Every paper will be allotted up to eight pages in the Proceedings. Camera-ready copy will be due to the Publications Chair, Zoubin Ghahramani, in early January Authors of accepted papers will be informed of the precise due date at the Conference. 8

11 NIPS*2001 REGISTRATION FORM We ask that you complete every entry on the registration form fully and legibly including a complete street and address. This information will be used to update the NIPS database. Note that there is a price reduction for registrations received by November 2, All prices on this page are in US dollars. Name Department Institution Name Street Address City State Zip/Postal Code Country Phone Fax Registration Fees: (Registration will not be accepted without payment in US dollars) Please check appropriate amounts: Tutorials: Regular $200 Full-time students $120 (with I.D.) Conference: After November 2, 2001 Regular $350 $400 Full-time students $210 $240 (with I.D.) Workshops: After November 2, 2001 Regular $250 $290 Full-time students $150 $175 (with I.D.) Bus from Vancouver to Whistler: $30 (Refundable only if paid by credit card) Proceedings: Hardcopy (Two volumes, soft cover) $30 (Special reduced rate available only if purchased before or during the Conference) Payment Method: Enclosed is a check or money order in US dollars for $ made payable to the Neural Information Processing Systems Conference. Bank wire transfer. (Wire transfers must be received by November 16) You may charge my credit card in the amount of $ MASTERCARD VISA Credit card number Name of credit card holder Expiration date Signature Workshop Preferences: Please list below the two Workshops that you are most likely to attend. This information will help us better organize the Workshop schedule. Workshop One Workshop Two 9