ConceptAI & ML
Policy gradient intuition
Policy gradients act as a weighted version of maximum likelihood where trajectories with high rewards have their log probabilities increased and trajectories with low rewards have their log probabilities decreased.
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.
Lecturers
1Connections
3The lectures that define it or assume it, and the concepts next to it.