Dynamically Adapting Kernels in Support Vector Machines
Nello Cristianini, Colin K. Campbell, John S. Shawe-Taylor · 1998
The kernel-parameter is one of the few tunable parameters in Support Vector machines, and controls the complexity of the resulting hypothesis. The choice of its value amounts to model selection, and is usually performed by means of a validation set. We present an algorithm which can automatically perform model selection and learning with no additional computational cost and with no need of a validation set. Theoretical results motivating this approach providing upper bounds on the generalisation error and experimental results confirming its validity are presented. Keywords: Support Vector Machine, Kernel-Adatron, Statistical Mechanics, Adaptive Kernels, Model Order Selection 1 Introduction Support Vector machines are learning systems designed to automatically deal with the accuracy/complexity trade-off, by minimizing an upper bound on the generalisation error provided by VC theory. In practice, however, SV machines still have a few tunable parameters which need to be set in order to ...