#decision-function

Question 1: Single choice

A trained SVM with the radial basis kernel $K(x, x') = \exp(-\lVert x - x' \rVert^2 / 2)$ predicts with $f(x) = \sum_{j \in SV} y_j \hat\alpha_j K(x_j, x) + \hat w_0$ . What happens to $f(x)$ at a test point $x$ that is very far from every training point?

It approaches 0, whatever the value of $\hat w_0$ 1
It approaches $y_j \hat\alpha_j$ of the nearest support vector2
It approaches the offset $\hat w_0$ 3
It grows without bound with the distance4

Timed test