Reward Deep Learning
noun
Definition: A scalar feedback signal produced by the environment in reinforcement learning, indicating the immediate value or desirability of an action outcome and serving as the basis for learning behavior that maximizes cumulative reward [Sutton, Barto 2017].
Example in context: “On taking an action, a reward is achieved and the overall objective is to maximize the sum of discounted rewards.” [Bai et al. 2023]
Related terms: reward signal, immediate reward