Chess engine with Convolutional Neural Networks

• Author: Adam Abed Abud
• Last update: September 9, 2021

Inspired and adapted from: https://github.com/victorsimrbt/chess_4096

Technology: Deep Learning, CNN, PyTorch

Introduction

This is a first, crude implementation of a chess engine based on CNN and implemented with pytorch.

The data used for the training is available from Kaggle: https://www.kaggle.com/datasnaek/chess

Prerequisites

Install the python chess module

sudo pip3 install python-chess

Instructions on how to run the code

First process the dataset games.csv by running

python generate_dataset.py

python process_dataset.py

Run the Jupyter-Notebook

jupyter-notebook pytorch_CNN_chess_engine.ipynb

Enjoy!

---

Future improvements

• Define better accuracy function to the model
• Increase the input layers in the deep learning process
• Improve the loss function

---

Project Organization

├── README.md                           <- README file for developers using this project.
|
├── games.csv                           <- Input file containing the training data
│
├── generate_dataset.py                 <- Convert input data into chess language
│
├── process_dataset.py                  <- Process the input data and make train and label datasets for the model
│
├── board_conversion.py                 <- Set of utilities for converting chess language into chess board
│
└── pytorch_CNN_chess_engine.ipynb      <- Main file responsible for the training, testing and visualization of the chess engine.

---

License

Free Software! For the benefit of everyone.