Interactive session, Monday April 24
There were unfortunately some technical issues with the recording of the first hour of the interactive session. The second hour is however linked in the schedule, and the slides from the first hour are found below.
Interactive session slides
Weekly lecture:
Slides:
Reinforcement learning (part 1)
Reinforcement Learning (part 2)
Videos:
- Introduction to Reinforcement Learning (RL)
- Policies and states in RL
- Rewards in RL
- Defining a policy in RL
- Learning algorithms in RL
- Q-learning example
- On-policy and off-policy learning
Readings:
Marsland Chapter 11
Optional Reading
The readings below go beyond the syllabus for this class, so do not worry if you don't understand everything. But, it can provide some interesting perspectives if you want to dig deeper into RL.
From Q-Learning to Deep Q-Learning: https://towardsdatascience.com/reinforcement-learning-tutorial-part-3-basic-deep-q-learning-186164c3bf4
The Paths Perspective on Value Learning: https://distill.pub/2019/paths-perspective-on-value-learning/
(Advanced) Article showing important limitations of state-of-the-art Reinforcement Learning: https://thegradient.pub/why-rl-is-flawed/