
Bounded Differences Inequality (Azuma-Hoeffding Inequality)
Yufei Zhao presents the bounded differences inequality, also known as the Azuma-Hoeffding inequality, in this lecture from MIT's 18.226 Probabilistic Methods in Combinatorics. The core result: a Lipschitz function of independent random inputs concentrates sharply around its mean, even when the function depends on many variables in complicated ways. Zhao develops the martingale-based proof technique behind the bound and then applies it to two classic problems, the coupon collector problem and the chromatic number of a random graph, showing how the same concentration argument yields precise tail estimates in both settings. The lecture assumes familiarity with basic probability and earlier course material on martingales. It is lecture nine of the course, continuing the term's focus on probabilistic tools for combinatorial problems, delivered chalkboard style with worked derivations rather than slides.