Action selection
Machine Learning
noun phrase
Definition: The process or decision problem of choosing the next action an intelligent agent should perform on the basis of its current state or observations in order to achieve its objective. In contemporary machine-learning usage, the term is most strongly associated with reinforcement learning, where action selection is governed by a policy and may be deterministic or stochastic [Maeda 2021].
Example in context: “The framework of RL with action selection and strategy update in Fig. 6 can already be used to solve some optimization problems, but has no enough consideration to the temporal changes in the system state and agent state.” [Zhou et al. 2024]
Related terms: policy decision, action-choice mechanism, exploration–exploitation strategy