California State University San Bernardino
Department of Computer Science and Engineering
Masters Project Presentation

Date
November 12, 2008
Time
1:30 pm ~ 2:30pm
Place
JB 389/391
Title
Implementation of Reinforcement Learning in Game Strategy Design

Candidate
Chien-Yu Lin

Advisor
Dr. Haiyan Qiao

Committee Members
Dr. Ernesto Gomez
Dr. David A. Turner

Abstract

Reinforcement learning is one type of machine learning. It is concerned 
with learning through the use of penalties and rewards. The main purpose of
this project is to apply reinforcement learning in the design of game 
strategy. Using this approach, the computer learns the opponent's strategy,
and learning takes place during each step of play. The reinforcement 
learning used in this project will be based on the Q-learning algorithm, 
and the game "Blackjack" is selected as the study model because of its 
simplicity and popularity.

In order to demonstrate that the strategy implemented with reinforcement 
learning performs better than pre-programmed strategies, an experimental 
approach is used. In the experiments, the winning percentages of different
strategies with and without learning capabilities are compared when playing
Blackjack.