An Introduction to Q-Learning Part 2/2

Hugging Face Blog · 4 years ago

Hugging Face published the second part of its free Deep Reinforcement Learning course, covering Q-Learning theory and implementation. The article explains that Q-Learning uses a Q-table initialized with 0 values for each state-action pair, updated using a temporal difference approach with a learning rate of 0.1 and discount factor of 0.99 in the provided examples. Learners are directed to implement a Q-Learning agent from scratch in two environments (Frozen Lake and a taxi navigation task) before progressing to Deep Q-Learning in the next unit.