
Why can a learning algorithm make reliable predictions on data it has never seen, even when trained on only finitely many examples? This book explores the mathematical principles that make such generalization possible. Beginning with the statistical learning model, the book develops...