Return Machine Learning
noun
Definition: The sum (or cumulative total) of rewards an agent expects to receive when following a policy over an episode (or from a state onward, depending on formulation) [Bakanach 2023/2024].
Example in context: After expansion, we have the rollout stage, where we select actions according to some default rollout policy for the remaining depth and compute the accumulated return. [Choudhury et al. 2022]
Related terms: cumulative reward, discounted return (specific form), reward-to-go