Consider the special case where we have a 1-dimensional feature vector and are interested in using a kernel rule. Suppose we have the training data (0,0),

(1,1), and (3,0), and we use the simple moving window classifier (i.e., with kernel function K(x) = 1 for |x| ≤ 1 and K(x) = 0 otherwise).

(a) Sketch the functions v0 3(x) and v1 3(x) for h = 0.5 and the classification rule. Indicate where there are ties. (Note the subscript simply denotes the fact that we have three training examples.)

(b) Repeat part

(a) for h = 1.

(c) Repeat part

(a) for h = 2.

(d) What happens for h ≥ 2.