Unlock: No-Regret Learning

Online learning against adversarial losses: regret as cumulative loss minus the best fixed action in hindsight, multiplicative weights, follow the regularized leader, and why no-regret dynamics converge to Nash equilibria in zero-sum games.