AI & MLRecommendedAIML-RL-008
Policy Optimization
Gradients through expected return: the REINFORCE idea and what it climbs.
- Difficulty
- 3 / 5
- Importance
- 4 / 5
- Study time
- 1–2 h
- Kind of work
- Mixed
Why it matters
The learner can narrate one policy-gradient step as upweighting actions from good episodes.