Glossary term
Glossary term
Agentic Systems
In reinforcement learning, each of the repeated attempts by the agent to learn an environment.
Created for this library
A robotics RL team groups simulated grasp attempts into episodes that begin with a new object placement and end when the robot drops or releases the object.
A trading research team treats each trading day as an episode in its RL setup so the agent learns within-day policies.
A game-AI team treats each match as an episode with terminal reward equal to win or loss for the trained agent.
Definition source: Google for Developers Machine Learning Glossary | Creative Commons Attribution 4.0 License