Action Masking
A technique in reinforcement learning that restricts the agent's set of selectable actions by masking the probabilities of illegal actions, ensuring that sampled actions are always legal and improving training efficiency and stability.
Timeline (last 90 days)
Action masking sets the probabilities of illegal actions to zero and renormalizes when the policy network outputs action probabilities, ensuring that the agent samples only from legal actions; it can significantly accelerate training but may introduce implicit preferences leading to path dependence.
The feasibility-aware action mask restricts the system to selecting only packets that can complete transmission within the remaining communication window.
All Facts (2)
Action masking sets the probabilities of illegal actions to zero and renormalizes when the policy network outputs action probabilities, ensuring that the agent samples only from legal actions; it can significantly accelerate training but may introduce implicit preferences leading to path dependence.
50%UnverifiedThe feasibility-aware action mask restricts the system to selecting only packets that can complete transmission within the remaining communication window.
50%