The package provides functions in order to smooth a set of curves. The idea is to perform non-parametric kernel regression of each of curves using the Nadaraya-Watson estimator. In order to use this estimator, we need to select a bandwidth. A common way to select this bandwidth is to used cross-validation. However, we are going to use the large number of curves to estimate it more efficiently. The idea is to estimate the underlying regularity of the curves (assuming they have the same). Then, the estimation of the bandwidth follows.