ICML / COLT 98 Overview A number of us were fortunate enough to be able to attend some of the many cool AI conferences that were co-located in Madison, WI this summer. In this meeting we will try to review some of what went on at the ICML and COLT conferences. The meeting will be organized as follows: We will start with a general open discussion of the big trends and the hot topics. Then a number of people will be making a series of short presentations about papers they found interesting: Andrew Moore - "Reinforcement Learning: How Far Can It Go?" by Rich Sutton envy.cs.umass.edu/~rich/ICML-98/index.htm Sean Slattery - "The Case Against Accuracy Estimation for Comparing Induction Algorithms" by Provost, Fawcett, and Kohavi www.croftj.net/~fawcett/papers/ICML98-final.ps.gz Kamal Nigam - "Combining Nearest Neighbor Classifiers through Multiple Feature Subsets" by Bay www.ics.uci.edu/~sbay/papers/mfs_icml98.ps Chuck Rosenberg - "An Efficient Boosting Algorithm for Combining Preferences" by Freund, Iyer, Schapire and, Singer www.research.att.com/~schapire/papers/FreundIyScSi98.ps.Z For more information about the talks given at ICML-98: http://www.cs.wisc.edu/icml98/schedule.html For more information about COLT-98: http://theory.lcs.mit.edu/COLT-98/