Geoffrey Hinton University of Toronto Canadian Institute for Advanced Research Overview of the tutorial • A brief history of deep learning. • How to learn multi-layer generative models of unlabelled data by learning one layer of features at a time. – What is really going on when we stack RBMs to form a deep belief net. • How to use generative models to make discriminative training methods work much better for classification and regression. • How to modify RBMs to deal with real-valued input.