Learning a Restricted Boltzmann Machine using biased Monte Carlo sampling