Feature learning is a set of methods that automatically discover useful representations from raw data for tasks such as classification, detection, or prediction. It reduces reliance on manually designed features and is central to many deep learning systems.