Algorithms for sequential decision-making are crucial in the operations of modern online platforms and web-based marketplaces. These algorithms should be designed to make better decisions as more information is revealed over time. The central theme of this course will be studying various families of these algorithms for online learning and prediction, and their applications to game theory and the design of electronic markets and platforms. The emphasis will be on mathematical techniques for designing, analyzing, and applying these algorithms. These mathematical techniques will include algorithms for predicting from expert advice, learning in games (including adaptive game playing, Nash equilibrium, and correlated equilibrium), stochastic and adversarial multi-armed bandit problems, contextual bandits and search, adversarial corruptions, incentivizing exploration, and Markov decision processes/reinforcement learning. We will learn these techniques through the lens of their applications to pricing and auction theory, automatic bidding, recommendation systems, calibrations in forecast, and design of clinical trials/experiments. The topic of each lecture is based on one particular application domain.
