Unlock: Multi-Armed Bandits Theory

The exploration-exploitation tradeoff formalized: K arms, regret as the cost of not knowing the best arm, and algorithms (UCB, Thompson sampling) that achieve near-optimal regret bounds.