Volume 2111 of the series Lecture Notes in Computer Science pp 224240
Rademacher and Gaussian Complexities: Risk Bounds and Structural Results
 Peter L. BartlettAffiliated withBIOwulf Technologies
 , Shahar MendelsonAffiliated withResearch School of Information Sciences and Engineering, Australian National University
Abstract
We investigate the use of certain datadependent estimates of the complexity of a function class, called Rademacher and gaussian complexities. In a decision theoretic setting, we prove general risk bounds in terms of these complexities. We consider function classes that can be expressed as combinations of functions from basis classes and show how the Rademacher and gaussian complexities of such a function class can be bounded in terms of the complexity of the basis classes.We give examples of the application of these techniques in finding datadependent risk bounds for decision trees, neural networks and support vector machines.
 Title
 Rademacher and Gaussian Complexities: Risk Bounds and Structural Results
 Book Title
 Computational Learning Theory
 Book Subtitle
 14th Annual Conference on Computational Learning Theory, COLT 2001 and 5th European Conference on Computational Learning Theory, EuroCOLT 2001 Amsterdam, The Netherlands, July 16–19, 2001 Proceedings
 Pages
 pp 224240
 Copyright
 2001
 DOI
 10.1007/3540445811_15
 Print ISBN
 9783540423430
 Online ISBN
 9783540445814
 Series Title
 Lecture Notes in Computer Science
 Series Volume
 2111
 Series ISSN
 03029743
 Publisher
 Springer Berlin Heidelberg
 Copyright Holder
 SpringerVerlag Berlin Heidelberg
 Additional Links
 Topics
 Industry Sectors
 eBook Packages
 Editors

 David Helmbold ^{(1)}
 Bob Williamson ^{(2)}
 Editor Affiliations

 1. School of Engineering, Department of Computer Science, University of California, Santa Cruz
 2. Research School of Information Sciences and Engineering Department of Telecommunications Engineering, Australian National University
 Authors

 Peter L. Bartlett ^{(5)}
 Shahar Mendelson ^{(6)}
 Author Affiliations

 5. BIOwulf Technologies, 2030 Addison Street, Suite 102, Berkeley, CA, 94704, USA
 6. Research School of Information Sciences and Engineering, Australian National University, Canberra, 0200, Australia
Continue reading...
To view the rest of this content please follow the download PDF link above.