Reinforcement Learning Algorithms

Custom implementations of Dynamic Programming, Monte Carlo, Temporal Difference and Vanilla Policy Gradient for learning purposes.

Algorithms Implemented

1. Dynamic Programming (DP):

2. Monte Carlo (MC):

3. Temporal Difference (TD):

4. Policy Gradients (PG):

Repository Structure

• Each algorithm has its dedicated folder (dynamic_programming/, monte_carlo/, temporal_difference/, policy_gradients/) containing code and related documentation.
• Each folder contains src/ for the source code, /notebooks for testing the code and for documentation. /notebooks is under construction. Temporarily the files in src/ can be ran directly to train and test the algorithms.
• _environments/ contains a modified verion of farama-foundations's FrozenLake environment that allows for modifying the reward structure.
• The code is designed to be readable and well-documented to aid in understanding and learning.