127 search results - page 6 / 26 » Training SVM with indefinite kernels |
CIARP
2010
Springer
13 years 5 months ago
2010
Springer
Abstract. It has been shown that many kernel methods can be equivalently formulated as minimal-enclosing-ball (MEB) problems in certain feature space. Exploiting this reduction eff...
NIPS
2004
13 years 9 months ago
2004
We describe an algorithm for support vector machines (SVM) that can be parallelized efficiently and scales to very large problems with hundreds of thousands of training vectors. I...
ICPR
2008
IEEE
14 years 2 months ago
2008
IEEE
The problem of signature verification is considered within the bounds of the kernel-based methodology of pattern recognition, more specifically, SVM principle of machine learning....
JMLR
2006
13 years 7 months ago
2006
While classical kernel-based learning algorithms are based on a single kernel, in practice it is often desirable to use multiple kernels. Lanckriet et al. (2004) considered conic ...
ICML
2008
IEEE
14 years 8 months ago
2008
IEEE
We propose a new stopping condition for a Support Vector Machine (SVM) solver which precisely reflects the objective of the Leave-OneOut error computation. The stopping condition ...