This is a fun project, inspired by talk of richard sutton - Tutorial: Introduction to Reinforcement Learning with Function Approximation

python3 learn_mdp.py

Here the user is a reinforcement learning agent and he tries to find the optimal policy to gain maximum rewards. The environment has two states A and B. User can take 2 actions - 1,2 . Based on user's action in a state he gets positive or negative reward/feedback.

If you decide to play this game then following is the optimal policy

This repository can be used for educational purposes. This repo can be used to explain the following concepts of Reinforcement Learning -

• MDP
• Exploration vs Exploitation Dilemma
• Introduction to RL.

Feel free to improve this project. Pull Requests are welcome.