P.E.A.R.L. is an open source DIY platform for understanding Reinforcement Learning. You learn at your own pace, through chapterwise slideshows, and then test your understanding on the corresponding coding assignments.
Slideshows are meant to provide better intuitions, using illustrations.
Feel free to raise issues or contribute to the course content. We continuously strive to improve the understandability and simplicity of the content, while keeping the completeness and intuitiveness of the foundational concepts.
The platform is a single paged application, with the course content present inside the "Episodes" section. Each episode is basically a seperate chapter. Why we named each chapter an "episode" is something you will figure out.
- Introduction
- Mathematical Treatment
- Evaluation Improvement Interplay
- Dynamic Programming
- Monte Carlo Methods
- Temporal Difference Learning
(More content to be updated regularly)
The assignments can be downloaded on clicking the "Code" button inside any episode. Assignments are in Python3. User will need to install the numpy and matplotlib libraries for proper execution. Assignments are designed in such a way that users need to fill in their code in the assigned area, and see the final output, and determine the accuracy and performance.