Abstract

This thesis explores the development of a personalized chess AI capable of mimicking an individual’s playstyle. The research delves into the evolution of chess AI, from early algorithmic engines to the current state-of-the-art neural network approaches. The study was conducted in two phases. The first phase employed traditional regression models to predict chess moves based on the author’s historical games, achieving moderate success. The second phase investigated the potential of neural networks, specifically the Maia chess engine designed to model human-like play. While fine-tuning Maia on the author’s game data encountered technical challenges, pre-trained Maia models demonstrated significantly higher accuracy in predicting the author’s moves compared to the regression model. This research contributes to the growing field of personalized AI, highlighting the potential for creating systems that adapt to individual behavior and preferences. Although limitations were encountered, the study paves the way for future explorations in personalized chess AI and its potential to enhance learning and enjoyment for players of all levels.

Advisor

Kelvey, Robert

Department

Computer Science

Disciplines

Artificial Intelligence and Robotics

Publication Date

2024

Degree Granted

Bachelor of Arts

Document Type

Senior Independent Study Thesis

Share

COinS
 

© Copyright 2024 Eltahir Elawad