The Robotics Institute

Carnegie Mellon Robotics Institute

Copyright notice: This material is presented to ensure timely dissemination of scholarly and technical work. Copyright and all rights therein are retained by authors or by other copyright holders. All persons copying this information are expected to adhere to the terms and constraints invoked by each author's copyright. These works may not be reposted without the explicit permission of the copyright holder.

Abstract

We propose a novel variant of conjugate gradient based on the Reproducing Kernel Hilbert Space (RKHS) inner product. An analysis of the algorithm suggests it enjoys better performance properties than standard iterative methods when applied to learning kernel machines. Experimental results for both classification and regression bear out the theoretical implications. We further address the dominant cost of the algorithm by reducing the complexity of RKHS function evaluations and inner products through the use of space-partitioning tree data-structures.

Keywords

kernel methods, functional gradient, reproducing kernel hilbert spaces, conjugate gradient, kernel logistic regression, regularized least squares, gaussian processes, kd-trees

Notes

Associated Center(s) / Consortia: Center for the Foundations of Robotics

Associated Lab(s) / Group(s): Planning and Autonomy Lab

Associated Project(s): Learning Locomotion

Number of pages: 8

Text Reference

Nathan Ratliff and J. Andrew (Drew) Bagnell, "Kernel Conjugate Gradient," tech. report CMU-RI-TR-05-30, Robotics Institute, Carnegie Mellon University, June, 2005

BibTeX Reference

@techreport{Ratliff_2005_5051,
   author = "Nathan Ratliff and J. Andrew (Drew) Bagnell",
   title = "Kernel Conjugate Gradient",
   booktitle = "",
   institution = "Robotics Institute",
   month = "June",
   year = "2005",
   number= "CMU-RI-TR-05-30",
   address= "Pittsburgh, PA",
}

The Robotics Institute

Menu