How agents learn by trial and error from rewards: environments, policies, value functions and the main algorithms.
55 terms
Thompson sampling is a Bayesian strategy in reinforcement learning that picks each action by sampling from its posterior belief about the action's reward.
Trust region policy optimization (TRPO) is a reinforcement learning algorithm that improves a policy while limiting each update with a KL divergence constraint.
Upper confidence bound (UCB) is a reinforcement learning strategy that picks the action with the highest optimistic estimate of its reward.
A value function in reinforcement learning estimates the expected discounted return an agent will collect from a state when it follows a given policy.
Value iteration is a dynamic programming algorithm in reinforcement learning that repeatedly applies the Bellman optimality update to find an optimal policy.