MDPs and Dynamic Programming · Online Reinforcement Learning

Skip to content

Topic

Online Reinforcement Learning

Multi-armed bandits and regret analysis

$ The Notion of RegretRegret bounds and online learning performance metrics $ Forms of Feedback and Problem SettingsFeedback models: full information vs. bandit $ Follow the Leader and HedgeFollow the Leader, Hedge, and expert algorithms $ Stochastic and Adversarial Bandits: UCB1 and EXP3UCB and EXP3 algorithms for bandit problems $ Contextual Bandits and EXP4Contextual bandits and the EXP4 algorithm $ MDPs and Dynamic ProgrammingMarkov decision processes and dynamic programming $ Monte Carlo Methods for RLMonte Carlo methods for policy evaluation $ Temporal-Difference Learning, SARSA, and Q-LearningTemporal difference, SARSA, and Q-learning $ Function Approximation and Deep Q-LearningDeep Q-networks and neural function approximation $ Online RL in Average-Reward and Discounted SettingsAverage reward criteria for continuing tasks

/topics/Online Reinforcement Learning/MDPs and Dynamic Programming

Loading...

Sign in to suggest