RL CH4 - Monte-Carlo Methods on Reinforcement Learning
In this Chapter:
Model-Based Learning and Model-Free Learning
First-Visit Method MC Prediction
Every-Visit Method MC Prediction
Monte Carlo Exploring Starts (MC-ES)
MC Epsilon Greedy (without Exploring Starts)
Aim of this chapter:
Understand the differences between model-based and model-free algorithms. Learning about Monte Carlo's different approaches both based on Prediction and Control.
*Note: The Rewards in the examples of these slides all are transition-based, not state-based (for simplicity of drawing and space they have shown in the states since all transition rewards to the same state are considered with the same values). The example's final results will be different if we consider state-based reward*