 Rademacher complexity

Rademacher complexity is a concept from statistical learning theory that measures the capacity of a class of functions or hypotheses in terms of their ability to fit random noise. Specifically, it quantifies how well a hypothesis class can "respond" to random labels.