In this recipe, we normalized the data at our disposal to the unitary norm. Each sample with at least one non-zero component was rescaled independently of other samples so that its norm was equal to one.
