Project - Dots and Boxes using Reinforcement learning

This project is an implementation of Dots and Boxes game using Reinforcement Learning - Q Learning Approach

• The project is implemented for 2X2 grid configuration and 3X3 grid configuration
• The Project can generate trained Q-table for games played between random agents
• This project can be used to view the results of Q-Agent played against a Random agent

Dots and Boxes game

Executing the project

2X2 Grid

Scripts in Src2 folder has the implementation of the game for 2X2 grid. Train the network by executing

python3 test.py

Dots and Boxes 2X2 game

3X3 Grid

Scripts in Src3 folder has the implementation of the game for 3x3 grid. Train the network by executing

python3 test.py

Dots and Boxes 3X3 game

Q-Learning Using DQN

The project is being implemented using Q- Learning with functional approximation. This part is not completed

python3 neural_network.py

Results

This project doesn't generate any visual results at this point of time.

The 2x2 Random Agent folder has the implementation of the game for 2x2 grid with a random agent. Use the following command to see the results

python3 test.py

The 3x3 Random Agent folder has the implementation of the game for 3x3 grid with a random agent. Use the following command to see the results

python3 test.py

All the results are included and updated in the report_1.pdf