Reinforcement Learning Crash Course - Deep Learning meets Reinforcement Learning | Reinforcement Learning Part-5

Video Created By - Rajtilak Pal (M. Tech in AI, IIT Ropar) Finally, we present to you the algorithm that changed the look of Reinforcement Learning. Deep Q-Learning or DQN (commonly referred to) is the algorithm that solved many difficult RL problems. In this video, we study what this algorithm is and code it out to see the results for ourselves in a couple of environments. Happy Learning! Intro Example Courtesy: Two Minute Papers - https://youtu.be/V1eYniJ0Rnk DQN 2013 Paper: https://arxiv.org/pdf/1312.5602.pdf DQN 2015 Paper from DeepMind: https://www.deepmind.com/publications/human-level-control-through-deep-reinforcement-learning CartPole Environment: https://www.gymlibrary.ml/environments/classic_control/cart_pole/ Mountain Car Environment: https://www.gymlibrary.ml/environments/classic_control/mountain_car/ Project Github Link: https://github.com/rajtilakls2510/reinforcement_learning/tree/part_5 Reinforcement Learning Book by Sutton and Barto: http://incompleteideas.net/book/the-book.html Free Reinforcement Learning Course from IIT Madras: https://nptel.ac.in/courses/106106143 ⌚Time Stamps⌚ 0:00 - Intro 0:43 - Recap problems from the previous video 1:08 - Elements in DQN 5:15 - Deep Q-Learning 6:11 - Implementing DQN 28:30 - Viewing Results of Our First Training Run 30:17 - Creating a new Reward function for CartPole 33:55 - Results for another training run 35:26 - More optimizations (what a surprise!) 43:07 - Writing Realtime Plots 46:47 - Training results after optimizing 49:51 - Mountain Car Environment 52:28 - Training Mountain Car 57:21 - Outro