The NIPS proceedings is widely held to be an archival journal quality publication. The high quality across such a diversity of topics is maintained through a rigorous review process. Out of 660 submissions received this year, the 196 papers presented here were selected by the program committee based on the recommendations of at least 3 reviewers per paper. The program committee members, emerging leaders in the field, each evaluated the papers and reviews in their area of expertise. Final decisions were made during an intensive two-day meeting of the Program Committee in Toronto, Ontario in which each paper was discussed and difficult cases were debated.
In recent years, the central themes of the NIPS meeting have been expanding beyond the boundaries of the traditional neural network model to include its close cousins such as Bayesian networks, MDPs, causal models, support vector machines and kernel methods. This year was no exception to this trend. Several exciting new themes have emerged over the last couple of years, including particle filtering (see papers by Thrun et al; Andrieu et al; and Fox), spectral clustering methods (Ng et al; Cristianini et al; Belkin & Niyogi; and Zha et al), active learning (e.g. Zimmerman et al), computing with spikes (e.g. Schwartz et al; Shon et al), and sparse coding in both real and artifical neural networks (e.g. O'Reilly & Busby; Fall & Tipping). Many truly unique and innovative pieces of work also appear in this volume, possibly defining the beginnings of trends for years to come. Among these, to name just a few, are Crammer & Singer's method for learning rankings of data, Yuille's ``Concave-Convex Principle'' for optimization, and Raphael's compelling musical accompaniment model. It is hoped that this admittedly sparse sample of the program's highlights will entice the reader to browse through the entire proceedings and discover its many hidden treasures.
Nominations for the NIPS best student paper award were considered by a subcommittee of the Program Committee. One winning paper and two papers for honourable mention were selected based on the strength of their contribution to the field, breadth of impact, and overall quality of presentation. The award was given to Evan Greensmith for his paper ``Variance Reduction Techniques for Gradient Estimates in Reinforcement Learning'' (E. Greensmith, P. L. Bartlett and J. Baxter). The authors who received honourable mention were Odelia Schwartz for her paper ``Characterizing neural gain control using spike-triggered covariance'' (O. Schwartz, E. J. Chichilnisky, and E. P. Simoncelli) and Martin Wainwright for his paper ``Tree-based reparameterization for approximate estimation in loopy graphs'' (M. Wainwright, T. Jaakkola, and A. Willsky).
At the meeting in Vancouver, the program of contributed papers was complemented by six enlightening invited talks. Following the opening banquet, the meeting's keynote address was delivered by Tom Knight (MIT) on ``Computing with Life'', providing a fascinating window into the new field of microbial engineering. Additional invited talks were given by Barbara Finlay (Cornell) on ``How brains evolve, and the consequences for computation'', Alison Gopnik (UC Berkeley) on ``Babies and Bayes-nets: Causal inference and theory-formation in children, chimps, scientists and computers'', Jon M. Kleinberg (Cornell) on ``Small-world phenomena and the dynamics of information'', Judea Pearl (UCLA) on ``Causal inference as an exercise in computational learning'', and Shihab Shamma (U Maryland) on ``Common principles in auditory and visual processing''.
The conference was preceded by a day of tutorials and followed by two days of workshops. The tutorials were organized this year by Yoshua Bengio, and presented by leading experts on a range of topics that spanned nonparametric density estimation (Luc Devroye), Bayesian networks (Daphne Koller and Nir Friedman), neurobiological modelling (Shawn Lockery), natural language modelling (Christopher Manning), SVM and Kernel methods (Bernhard Schölkopf), and robotics (Sebastian Thrun). The workshop program included a wide array of parallel sessions, permitting more casual, in-depth presentations and discussions, and was organized with tremendous success this year by Virginia de Sa and Barak Pearlmutter.
The software for electronic submissions of both papers and reviews, now in its third year of operation, and NIPS web pages, ran smoothly thanks to the efforts of Xin Wang (Oregon State University), Alexander Gray (Carnegie Mellon University), and Phil Galbiati (Oregon Graduate Institute).
Throughout the organization of this conference, we were continually impressed by the enthusiastic commitment of the many people who contributed their efforts to its success. First and foremost are the authors, whose hard work has produced the excellent papers contained in this volume. Second, we thank the thirteen members of the Program Committee who applied their knowledge and experience to make the difficult acceptance decisions. The program committe members operated as a phenomenal team in dealing with 660 submissions and nearly 2000 reviews -- an overwhelming 30% increase in volume above the previous few years. Third, we thank the members of the Organizing Committee including the workshops co-chairs Virginia de Sa and Barak Pearlmutter, the tutorials chair Yoshua Bengio, the publicity chair Richard Zemel, the treasurer Bartlett Mel, the volunteers director Sidney Fels, and the local arrangements chair Rosemary Miller. Yoshua Bengio assembled an outstanding set of tutorial speakers. The workshop program co-chaired by Virginia de Sa and Barak Pearlmutter covered a diverse and exciting range of topics. The wonderful ``Moon-brain over Vancouver'' image on the NIPS publicity poster was the brainchild of Rich Zemel (Publicity Chair), and was produced by Linda Gustafson of Counterpunch, Inc., Toronto. Rosemary Miller performed the lion's share of the organizational duties, and deserves thanks and credit for the seamless transition of the NIPS conference, after 14 years in Colorado, to Vancouver and Whistler B.C.
Special thanks are due to Xin Wang, who did a substantial amount of proramming to enable secure credit card transactions in the online registration process, and to Alex Gray for substantial improvements to the appearance of the NIPS web page.
We are grateful for all the work Chris Togias did to help produce the conference abstract book and proceedings, and for the fabulous design of the abstract booklet cover by Monica Biagioli. We also thank the student volunteers who helped with conference logistics, registration, and the poster and workshops sessions.
Finally, we thank the NIPS Foundation Board, whose work over the years has contributed to the development of a richly interdisciplinary community with high standards of excellence and dedication.
Thomas G. Dietterich, Oregon State University
Sue Becker, McMaster University
Zoubin Ghahramani, University College London
December 2001