Deep learning and computer chess: Giraffe

• Main Author: Chan, Eu Ching
• Other Authors: He Ying
• Format: Final Year Project
• Language: English
• Published: Nanyang Technological University 2024
• Subjects: Computer and Information Science
• Online Access: https://hdl.handle.net/10356/175086
• Institution: Nanyang Technological University

This study aims to bridge the gap between advanced machine learning techniques and strategic game theory, embracing the transformative impact of Artificial Intelligence in the world of chess. It focuses on building a deep-learning chess engine, Giraffe, which is the first-ever neural network model to evaluate positions by exploring the application of neural network techniques in evaluating chess positions, digging into the network architecture, and training methodologies using extensive game datasets. The trained evaluation network then acts as an early playout termination in Monte-Carlo Tree Search (MCTS). The report highlights the engine’s ability to discover all its domain-specific knowledge, with minimal information given by the dataset. By conducting a 100-game match with each engine, this report will demonstrate the effectiveness of Giraffe through comparative analysis with traditional evaluation algorithms.