Introduction to Artificial Intelligence: Temporal Difference Learning - Lecture 17

Explore key concepts in artificial intelligence through this comprehensive lecture on temporal difference learning and related topics. Delve into on-policy vs off-policy methods, soft policies, and Monte Carlo techniques. Examine temporal difference learning, including the TD(0) algorithm, and compare it to Monte Carlo methods using real-world examples. Investigate TD control methods like SARSA and Q-Learning, analyzing their performance in scenarios such as "The Cliff." Review potential exam questions and get an overview of Assignment 5, including GUI demonstrations and code structure explanations for Maps.js, Environment.js, and RL_Student.js. Gain valuable insights into modern AI problem-solving techniques applicable to game development and beyond.

- Preroll
- Greetings
- Lecture Begin
- On-Policy vs Off-Policy
- Soft Policies
- On-Policy First-Visit MC soft
- Example Epsilon-Soft Calculation
- Off-Policy Methods
- Temporal Difference Learning
- TD0 Algorithm
- MC vs TD Example: Driving Home
- Advantages of TD
- TD Control
- SARSA
- Q-Learning
- The Cliff: SARSA vs Q-Learning
- Exam Questions
- Assignment 5 Overview / GUI
- Example Movement / Update Step
- Code Overview
- Maps.js
- Environment.js
- RL_Student.js