Jack ChenTJHSST Computer Systems Lab 2009-2010

AbstractThe purpose of this project is to explore Artificial Intelligence techniques in the board game Othello. The project focuses on the creation and evaluation of AI players for Othello. It explores several techniques used in strong AI players, including improvements to minimax game-tree search algorithms and higher-quality evaluation functions. It also investigates machine learning methods to enable AI players to improve the quality and speed of play automatically based on training and experience.

Scope of StudyRefereeGUIAI players

Search algorithmsBoard evaluation functionsMachine learning

RefereeWritten in PythonSupports varying types of players

AI programs in multiple languagesHumans

Runs the players in child processesCan display board in a GUI

GUIWritten in C++Qt graphics toolkitDisplays boardAnimates players'

movesAllows human to

play easily

Search AlgorithmsMinimaxAlpha-beta pruningNegaScoutMTD(f)Selective searchQuiescence search

Board Evaluation FunctionComplex features: mobility, stabilitySimple features: patterns of disksTraining with machine learning

Stochastic gradient descentGenetic algorithmsNeural networks, backpropogation

Other AI enhancementsBitboardsTransposition tablesOpening bookParallelization

Expected ResultsEvaluate players based on performance in

games against other playersCompare several different AIsPlayers using machine learning would be

expected to improve