|
ACM HONORS CREATORS OF NEW BOOSTING ALGORITHM
AdaBoost Improves Accuracy in Machine Learning
|
NEW YORK, March 15, 2005 -- ACM has recognized two leading theoretical scientists for their contribution to highly accurate prediction rules used in a range of machine learning applications associated with artificial intelligence. Yoav Freund of Columbia University and Robert Schapire of Princeton University introduced a new boosting algorithm called AdaBoost which can be used to significantly reduce the error of algorithms used in statistical analysis, spam filtering, fraud detection, optical character recognition, and market segmentation, among other applications. Drs. Freund and Schapire will receive the Paris Kanellakis Theory and Practice Award, which carries a $5,000 prize.
Dr. Schapire's breakthrough 1990 paper proved that boosting, a way to produce very accurate prediction rules used for classification tasks, could be used to improve prediction accuracy by combining "weak" prediction rules, (i.e. simple "rules of thumb" whose accuracy is only slightly better than random guessing). A year later, Dr. Freund developed a more efficient boosting algorithm. In 1995, their collaboration resulted in AdaBoost (short for Adaptive Boosting), an efficient, simple and easy-to-program learning strategy. In combination with other learning methods, AdaBoost has been described as "the best off-the-shelf classifier in the world." AdaBoost's accuracy is achieved by combining many weak "rules of thumb", each one a bit better than random guessing, to arrive at a single combined prediction rule that can be highly accurate.
The AdaBoost Algorithm's elegance, wide applicability, simplicity of implementation and great success in practice has transformed boosting into one of the pillars of machine learning. The algorithm is widely used and continues to expand in its relevance and importance to the practice of machine learning. For example, it has been used in spoken dialogue systems to categorize human utterances according to their meaning. In computer vision it is used to find all the human faces in an image. In computational biology, it is used to identify proteins that regulate gene expression. In fraud detection, it is employed to identify fraudulent credit card or calling card transactions. In optical character recognition, it is applied in reading zip codes or handwritten bank checks. In market segmentation, it predicts if a customer will respond to a marketing promotion.
Dr. Freund attended Hebrew University in Israel, where he received a B.Sc. in physics and math in 1982, and a M.Sc. in computer science under the supervision of Eli Shamir in 1989. He earned a Ph.D. in Computer Science from the University of California, Santa Cruz in 1993, under the supervision of Manfred Warmuth. From 1993 to 2001 he served as a member of the technical staff in AT&T Labs (formerly AT&T Bell Laboratories). Following a year at Banter, Inc., a startup company, he joined the Center for Computational Learning Systems at Columbia University as a senior research scientist.
Robert Schapire received his Sc.B. in math and computer science from Brown University in 1986, and his M.S. (1988) and Ph.D. (1991) from MIT under the supervision of Ronald Rivest. After a post-doctoral program at Harvard University, he joined the technical staff at AT&T Labs (formerly AT&T Bell Laboratories) in 1991. In 2002, Dr. Schapire became a Professor of Computer Science at Princeton University. He was awarded the 1991 ACM Doctoral Dissertation Award.
In 2003, Drs. Freund and Schapire received the Goedel Prize for their work on the AdaBoost learning algorithm.
ACM will present the 2004 Kanellakis Award to the winners at its annual Awards Banquet on June 11, in San Francisco, CA. The Paris Kanellakis Theory and Practice Award honors specific theoretical accomplishments that have had a significant and demonstrable effect on the practice of computing. This award is endowed by contributions from the Kanellakis family, with additional financial support provided by ACM's Special Interest Groups on Algorithms and Computational Theory (SIGACT), Special Interest Group on Design Automaton (SIGDA), Special Interest Group on Management of Data (SIGMOD), Special Interest Group on Programming Languages (SIGPLAN), the ACM SIG Discretionary Fund, and individual contributions.
About ACM
ACM (www.acm.org) is widely recognized as the
premier organization for computing professionals, delivering a broad array
of resources that advance the computing and IT disciplines, enable
professional development, and promote policies and research that benefit
society. For further information, contact Virginia Gold 212-626-0505 or vgold@acm.org.
|
|
|
|
|
