4. Non-parsimonious core methods

The correct use of kernels is linked to a representation technique that enables us to move from an initial "functional" formulation (where the space of hypotheses is a set of functions of type $\mathcal{H}$ ) to a second formulation, this time vector-based, showing, for each example, a coefficient representing the influence of this point in the solution. To illustrate this principle, let's take the example of interpolation splines.