Computers Attempting to Kick Human Butt at Connect 4

About

This is the final project for CS 467: Machine Learning by JimmyJHickey and BAndrewss.

We created a Connect 4 game. There are four different types of players that can participate: Human, Random, and Minimax and Neural Network players to play it.

Starting the Game

• Have Python 3 installed.

• Navigate to the /src directory in a terminal or cmd.

• Run the game.py file with 2 arguments.

  • python3 game.py *arg1 arg2*

• These arguments specify which type of player will be player 1 and 2 respectively. These are the options:

  • human - a human player
  • random - a random player
  • mini_easy - a minimax player of depth 3
  • mini_medium - a minimax player of depth 5
  • mini_hard - a minimax player of depth 6
  • net_random - neural network player trained against random data
  • net_easy - neural network player consisting of one layer of 4 nodes
  • net_medium - neural network player consisting of 100 | 100 nodes
  • net_hard - neural network player consisting of 1000 | 1000 nodes

---

About the project

The Python Files

• game.py contains the game and game board logic.
• players.py contains all of the different player base logic.
• supervised_net.py contains all of the neural network training logic.
• minimax.py contains all of the minimax logic.
• settings.py contains all of the shared data for players and game.

The Data

All of the data that was used to train neural networks is stored in /data.

All of the trained neural network data is stored in /trained_networks.

Reports

Both the proposal and final paper were written in LaTeX. The TeX and the PDFs that were created are in /Reports.

The Jupyter Notebook used for the presentation is in /src so it works correctly with our Python. It is best viewed on the Github. (www.github.com/JimmyJHickey/Machine-Learning-Connect-Four)