Representation Learning

The process by which a model automatically learns useful ways to represent raw data, such as turning pixels into meaningful visual features or words into semantic vectors. Good representations capture the structure that matters for a task and discard irrelevant noise.