RL is very mathematical. It is advisable to keep the following prereqs ready before starting out with RL. (Look at the respective topic.md files in Slack-Stock-DAG):
- Python
- Calculus
- Probability and Statistics
- Linear Algebra
- ML and DL
- PyTorch (or Tensorflow)
- Data Structures and Algorithms
Here's the Pathway for RL:
- Watch David Silver's course on RL basics.
- Read the first 7 chapters of Hands on Deep Reinforcement Learning to get an idea of how DL and RL meet and how DQNs work and also some nice exposure to code.
- Deep RL Bootcamp and CS285 are the most important courses for RL (in 2020). Use them for building up a strong theoretical background.
- At any point you want to get deeper into any classic RL algorithm, dig deeper into the respective chapter in Sutton Barto 2e
- When you want to practice DQNs and PGs, you can take help from these implementations: Rainbow is all you need and PG is all you need.
- Read a lot of papers, try out own experiments, etc. Spinning Up should help a lot with this.