Groups
Category
Concentration inequalities give high-probability bounds that random outcomes stay close to their expectations, even without knowing the full distribution.
Statistical learning theory explains why a model that fits training data can still predict well on unseen data by relating true risk to empirical risk plus a complexity term.