In contrast to Supervised Learning, we don’t have a ground truth, and the data is unlabeled. We therefore need to identify structures in the data.