Reinforcement Learning Explained
Agents learn to maximize long-term rewards through interaction with an environment modeled as a Markov Decision Process.
📚 Concepts
- Value function: expected return from a state
- Bellman equations: recursive relationships
- Exploration vs exploitation: balance to learn effectively
❓ Frequently Asked Questions
1) Model-based vs model-free?
Learn dynamics vs learning values/policies directly.
Learn dynamics vs learning values/policies directly.
2) On-policy vs off-policy?
Whether learning targets the behavior policy or a different one.
Whether learning targets the behavior policy or a different one.
3) Function approximation?
Neural networks approximate value/policy for large spaces.
Neural networks approximate value/policy for large spaces.
4) Stability?
Target networks and experience replay improve stability.
Target networks and experience replay improve stability.