Online Learning: Regret Minimization, the Multiplicative Weights Algorithm, and Adversarial BanditsA rigorous treatment of online learning—regret minimization, multiplicative weights, EXP3 for adversarial bandits, and the deep connections to game theory and boosting.
