As outlined in Section 5.3.1, smoothing and filtering can be a useful tool for preprocessing data. We consider Friedman's super-smoother [Fri84] in domains where it is appropriate (i.e. high-noise).
