Unsupervised Feature Learning
& Deep Learning

Machine learning has seen numerous successes, but applying learning algorithms today often means spending a long time hand-engineering the input feature representation. This is true for many problems in vision, audio, NLP, robotics, and other areas. To address this, we are developing unsupervised feature learning algorithms that automatically learn a good representation for the input. Since these algorithms mostly learn from unlabeled data, they have the potential to learn from vastly increased amounts of data (since unlabeled data is cheap), and therefore also achieving vastly improved performance.