What you'll learn

Monte Carlo methods can feel abstract, leaving practitioners unsure how to turn episodic returns into effective value estimates, policies, and trainable networks. In this course, Monte Carlo Methods, you’ll learn to build and evaluate Monte Carlo-based reinforcement learning agents end to end. First, you’ll explore Monte Carlo prediction with episodic sampling and the differences between first-visit and every-visit estimation. Next, you’ll discover Monte Carlo control using ε-greedy policies to derive optimal behavior from experience. Finally, you’ll learn how to implement the REINFORCE policy-gradient algorithm in PyTorch and assess its performance on CartPole. When you’re finished with this course, you’ll have the skills and knowledge of Monte Carlo methods in reinforcement learning needed to design, implement, and evaluate prediction, control, and policy-gradient agents.