A hype of machine learning based on artificial neural networks in which multiple layers of processing are used to extract progressively higher-level features from data. The adjective "deep" refers to the use of multiple layers in the network. Methods used can be either supervised, semi-supervised, or unsupervised.