Epsilon-Greedy Strategy
The most commonly used exploration-exploitation balancing strategy in reinforcement learning, in which the agent explores randomly with probability epsilon and exploits the current best action with probability 1 minus epsilon.
Timeline (last 90 days)
In thehumanbox, action selection prioritizes survival needs and then executes an ε-greedy policy based on the learned values.
In the ε-greedy policy, the agent chooses an action at random with probability ε and chooses the action with the highest current estimated Q-value with probability 1-ε.
In a civilization sandbox, pure ε-greedy random exploration is extremely inefficient when the action space is about 5000, and hand-written priorities can greatly accelerate early learning.
Common methods for addressing the exploration-exploitation trade-off include epsilon-greedy strategies, optimistic initialization, and UCB.
All Facts (4)
In thehumanbox, action selection prioritizes survival needs and then executes an ε-greedy policy based on the learned values.
50%UnverifiedIn the ε-greedy policy, the agent chooses an action at random with probability ε and chooses the action with the highest current estimated Q-value with probability 1-ε.
50%UnverifiedIn a civilization sandbox, pure ε-greedy random exploration is extremely inefficient when the action space is about 5000, and hand-written priorities can greatly accelerate early learning.
50%UnverifiedCommon methods for addressing the exploration-exploitation trade-off include epsilon-greedy strategies, optimistic initialization, and UCB.
50%