Finding the correlations and magnetizations of a Ising model from a set of couplings is a hard, but well studied, problem. In the other hand, the inverse problem of finding the couplings from a set of measured data has only recently got the attention of the scientific community. In this talk I will present some analytical results for a particular case of this problem: infering the patterns of a Hopfield model. This approach has the advantage of yielding an analytical solution of the patterns in the large size limit, but also allows for explicit calculations using replica methods of how much data is required for an accurate estimation