This is a series of companion videos to Sutton & Barto's textbook on reinforcement learning used by some of the best universities as standard course text - including Stanford, UCL, Carnegie Mellon
Download Book for free: incompleteideas...
Buy on Amazon: www.amazon.com...
Github Code Repository: github.com/Sha...
This first video covers the first few sections of chapter one.
00:00:00 Video intro
00:00:35 Why follow Sutton & Barto's Reinforcement Learning Textbook
00:00:50 Where to download the book for free
00:01:30 Reinforcement Learning in Humans and Animals (David Silver's UCL course slide)
00:02:00 Motivations for learning reinforcement learning and importance for real life problems
00:02:30 Personalisation for marketing and online
00:02:43 Control systems in commercial climate control
00:02:55 ChatGPT & Reinforcement Learning with Human Feedback (RLHF)
00:03:10 Google Deepmind AlphaGo Zero for superhuman capability
00:04:05 RL as a type of problem and as a set of tools
00:04:20 Supervised Learning vs. Unsupervised Learning vs. Reinforcement Learning
00:06:17 Reinforcement Learning vs. Artificial Neural Networks
00:07:00 Key characteristics of reinforcement learning problems
00:07:10 Example: Pavlova vs. Mochi - Nemesis
00:07:22 Mr. Stick: Rewards and Action set
00:07:55 Pavlova's goal - as many treats as possible
00:08:15 Pavlova's environmental state
00:08:40 Stochasticity of environment
00:09:29 Pavlova's policy
00:10:00 Trial and error search for rewards
00:10:52 4 key characteristics of RL problem: goal, state, actions and sequence
00:11:30 Key components of an RL solution: Policy, Reward Signal, Value Function, Model