Aerospace applications of Gaussian processes, Hilbert spaces and wavelets

The problem of learning from finite, noisy data sets is ill-posed in the sense that a solution may not exist, be unique or depend continuously on the data. The classical way to solve this learning problem is regularisation theory. This is described and shown to be interpretable in Hilbert spaces, as Gaussian process priors or in terms of frequency domain characteristics. Some remarks are also given regarding a connection with approximation by wavelets