ConceptAI & ML
Epsilon-greedy policy improvement
A policy improvement step where a greedy policy is modified to choose a random action with probability epsilon, guaranteeing an improvement over the original value function.
Know first
Named as a prerequisite by the lectures that define this. The number is how many say so. A grey one has no page yet.
Action-value function1Policy improvement theorem1
Lecturers
1Connections
2The lectures that define it or assume it, and the concepts next to it.