Deep learning and computer chess
This report presents two supervised learning approach for training neural networks to evaluate chess positions. The architecture used to build the neural network model is based on the Giraffe’s architecture [2] and Stockfish NNUE -HalfKP [3]. Implemented a method to train a neural network architectu...
Saved in:
| Main Author: | |
|---|---|
| Other Authors: | |
| Format: | Final Year Project |
| Language: | English |
| Published: |
Nanyang Technological University
2022
|
| Subjects: | |
| Online Access: | https://hdl.handle.net/10356/162874 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Institution: | Nanyang Technological University |
| Language: | English |
| Summary: | This report presents two supervised learning approach for training neural networks to evaluate chess positions. The architecture used to build the neural network model is based on the Giraffe’s architecture [2] and Stockfish NNUE -HalfKP [3]. Implemented a method to train a neural network architecture to understand chess movement and techniques that a grandmaster would play. Both approaches implemented as a 7-class classification problem on a dataset of over 10,000 samples games.
We collected different chess game played by grandmaster, then used the evaluation function of stockfish [5], one of the strongest existing chess engines, to get the score of the positions and label it accordingly. We extracted the positions from the games using Forsyth-Edwards notation and stored them in csv files which are later used for training the model. |
|---|