Introduction to (deep) reinforcement learning
summary
An accessible introduction to reinforcement learning covering its mathematical and algorithmic foundations and why modern methods work. Includes Python implementations and examples, and a brief discussion of function approximation using deep neural networks (deep reinforcement learning), with references to recent successes like DeepMind’s Atari and Go agents.
AI-written summary · about
chapters (28)
- 0:00Introduction
- 0:22What is AI?
- 1:05Machine learning origins
- 3:12Reinforcement learning
- 6:24Inverted pendulum demo
- 9:42RL difficulty
- 11:06Results after 20k episodes
- 12:55Pac-Man common mistakes
- 14:39Dimensionality reduction
- 16:26RL formal definition
- 17:09Bellman equation
- 18:16Markov processes
- 19:38MDPs
- 21:01Value function and reward
- 22:26Dynamic programming
- 23:51Model-free Q-learning
- 25:15Q-learning algorithm
- 26:18Exploration vs exploitation
- 27:01Pixel challenge
- 28:04Deep Q-learning
- 29:07Breakout video
- 31:20Conclusions and future
- 32:08Multiagent questions
- 35:26Q-learning convergence
- 37:13Math prerequisites
- 39:16Libraries and tools
- 40:20Neural network size
- 42:31Closing and Q&A
related talks
44mReinforcement learning: A gentle introduction
05mPractical insights into deep reinforcement learning
49mReinforcement Learning: Demystifying the hype
01h 21mIntroduction to reinforcement learning using Python & OpenAI Gym
28mDeep Reinforcement Learning: theory, intuition, code
01h 17mTraining intelligent game agents using deep reinforcement learning
30mReinforcement Learning - ChatGPT, Playing Games & More
49mML: Teaching machine to learn from its own mistakes (...)