Looking Beyond Data: Information-theory based criteria to evaluate different machine-learning approaches for hydrological problems