Deep learning approaches, leveraging on big data, are outperforming state-of-the-art more “classical” supervised and unsupervised approaches, directly learning relevant features and data representations without requiring explicit domain knowledge or human feature engineering. These approaches are currently highly important in IoT applications.