CS DISTINGUISHED ALUMNI LECTURE SERIES

Tom Mitchell

Does Machine Learning Really Work?

September 19, 1996

4:00 pm, Wean Hall 7500


ABSTRACT
Yes. The past decade has seen rapid progress on understanding how to make machines learn. In ten years we have gone from algorithms that were laboratory curiosities to robust methods with significant commercial value. Machine Learning algorithms now learn to control vehicles to drive autonomously on public highways at 70 mph, learn to detect credit card fraud by mining data on past transactions, and learn your reading interests in order to assemble a personally customized electronic newspaper. At the same time, new theoretical results shed light on fundamental issues such as the tradeoff among the number of training examples available, the number of hypotheses considered, and the likely accuracy of the learned hypothesis. And work on integrated learning architectures is beginning to explore issues such as long-term learning of new representations, cummulative learning, and learning to learn.

Where is all this headed? This talk will examine recent progress and open questions in machine learning, suggest some research topics that we should begin on now, and give one person's view on where machine learning might be headed over the next decade.

SPEAKER BIO
Tom Mitchell is Professor of Computer Science and Robotics at Carnegie Mellon University. He earned his B.S. degree from MIT (1973), and his Ph.D. degree from Stanford University (1979). In 1983 he received the IJCAI Computers and Thought award in recognition of his research in machine learning, and has been a Fellow of the American Association for Artificial Intelligence since 1990. In 1995 he helped found Schenley Park Research, Inc., a CMU spinoff specializing in commerical applications of machine learning and data mining. His new textbook "Machine Learning" will be published by McGraw Hill in January 1997. Mitchell's current research focuses on new algorithms that combine prior knowledge with observed data to improve learning accuracy, on algorithms for learning over multimedia data (e.g., combined symbolic, numeric, text, images), and applications of machine learning to web-based agents, robotics, and data-mining. But his real goal is to apply it to windsurfing. Mitchell's address is School of Computer Science, 5000 Forbes Ave., Carnegie Mellon University, Pittsburgh, PA 15213 (Tom.Mitchell@cmu.edu, http://www.cs.cmu.edu/~tom).

SCS Distinguished Lecture Schedule
SCS Calendar of Events