exploration

Definition

Deliberately trying actions whose outcomes are still uncertain, in order to gather information. Without it an agent cannot discover that a better option exists, and it locks onto whatever looked good early — a policy that is locally fine and globally poor. Exploration is the counterpart of exploitation, and balancing the two is one of the defining design problems of reinforcement learning, addressed through epsilon-greedy schedules, entropy bonuses, and intrinsic-reward signals.

Explainers using this term

• Paper Walkthrough — WarpSAC: When RL's Safety Rails Become HandcuffsWarpSAC: Towards the Pinnacle of Scalable Off-policy RL by Rethinking Exploration and Exploitation