Practical aspects of using core machines
Kernel methods for statistical learning

In the parametric framework, statistical studies begin with a model definition phase, followed by a parameter identification phase and ending with a validation study. In the non-parametric framework, i.e. using a kernel machine, the modeling phase is replaced by a rather complex methodology in which the kernel must be chosen and the hyperparameters identified, which amounts to selecting a suitable model. This phase is often very costly in terms of computing time, hence the need for an efficient method for parameter identification (solving the problem posed by the equations (8) ).

We will present various aspects related...