Session XVI: Deep Reinforcement Learning (Policy Gradients and SLM-Lab)
Meeting date: February 17th, 2018
For this session, Laura Graesser and Wah Loon Keng built on the Deep Reinforcement Learning foundations from their December session at the Deep Learning Study Group. This time, they focused on Policy Gradient algorithms, specifically REINFORCE (slides here) as well as their SLM-Lab Python library, which replaces their OpenAI Lab for experimenting with and optimising Deep RL agents across a wide variety of environments.
A summary blog post, replete with photos of the session, can be found here.
Recommended Preparatory Work
- Install SLM-Lab and bring along your laptop if you'd like to experience Keng and Laura's demo interactively.
- CS231n Spring 2017 Lecture 14 by Serena Young on Deep Reinforcement Learning (my notes on the lecture are here)
- In addition, you may enjoy any of the following previously-recommended resources if you haven't perused them already:
Laura will cover additional Deep Reinforcement Learning topics, like memory, initialisation and the Actor-Critic algorithm.