Playing Tic Tac Toe with Neural Networks Justin Herbrand CS/ECE/ME 539
Reason Entertainment Learn the basic idea behind other video games AI Learn what is going on behind the scenes of video games
Goal Build a game of tic tac toe that could play a human player Build a game of tic tac toe that could play a human player Try predict moves where the human player might be going and defend against it Try predict moves where the human player might be going and defend against it Use a MLP to predict the nonlinear pattern that human players have Use a MLP to predict the nonlinear pattern that human players have
Why MLP It can predict non-linear patterns since not to many people follow a specific pattern. It can predict non-linear patterns since not to many people follow a specific pattern. See if we can teach a computer to play games and not just give it a algorithm to play it. See if we can teach a computer to play games and not just give it a algorithm to play it.
Results Have a working game where the human can interact with the computer Have a working game where the human can interact with the computer Not a very difficult neural network that challenges the human player Not a very difficult neural network that challenges the human player
Game Play Basically you pick a number that corresponds to the locations of the board to make a move Basically you pick a number that corresponds to the locations of the board to make a move Then the computer goes Then the computer goes Goal: Get three in a row Goal: Get three in a row
Development Try to improve the neural network to make the game harder for the user to play Try to improve the neural network to make the game harder for the user to play Develop better data points to train the MLP what a good move will be Develop better data points to train the MLP what a good move will be