Reinforcement Learning: ARENA

I spent some time as both a student and a teaching assisstant at the second edition of ARENA, an AI Alignment upskilling workshop running in London, June 2023.

I gave some talks on Reinforcement Learning, covering the basics (bandits, MDPs, tabular learning, SARSA, Q-Learning) and some exploration of other RL methods such as DQN and VPG. Here are the slides (handout and present) and the corresponding tex source.

NB: don’t trust the PPO material, I didn’t end up using it, but it’s included for posterity, and will likely change for future iterations.

---