public GameForm() { InitializeComponent(); _randomPlayer = new RandomPlayer(); _optimalPlayer = new OptimalPlayer(SquareTypes.O); _neatPlayer = new NeatPlayer(null, SquareTypes.O); _aiSquareType = SquareTypes.O; _humanSquareType = SquareTypes.X; _game = new TicTacToeGame(); // Set the AI to the random player by default. _ai = _randomPlayer; // Experiment classes encapsulate much of the nuts and bolts of setting up a NEAT search. _experiment = new TicTacToeExperiment(); _hyperNeatExperiment = new TicTacToeHyperNeatExperiment(); // Load config XML for the NEAT experiment. XmlDocument xmlConfig = new XmlDocument(); xmlConfig.Load("tictactoe.config.xml"); _experiment.Initialize("TicTacToe", xmlConfig.DocumentElement); // Load config XML for the HyperNEAT experiment. xmlConfig = new XmlDocument(); xmlConfig.Load("hyperneat.config.xml"); _hyperNeatExperiment.Initialize("TicTacToe", xmlConfig.DocumentElement); }
private static void MatchReplay() { int inputCount = 9; int outputCount = 9; int depth = 3; int hiddenNeuronsPerLayer = 5; NeuralPlayer champion = MakeNeuralPlayer(inputCount, outputCount, depth, hiddenNeuronsPerLayer, new DNA(bestDNA)); OptimalPlayer p2 = new OptimalPlayer(SquareTypes.O); champion.SquareType = SquareTypes.X; TicTacToeGame.PlayGameToEnd(champion, p2); }
/// <summary> /// Evaluate the provided IBlackBox against the random tic-tac-toe player and return its fitness score. /// Each network plays 10 games against the random player and two games against the expert player. /// Half of the games are played as circle and half are played as x. /// /// A win is worth 10 points, a draw is worth 1 point, and a loss is worth 0 points. /// </summary> public FitnessInfo Evaluate(IBlackBox box) { double fitness = 0; SquareTypes winner; OptimalPlayer optimalPlayer = new OptimalPlayer(SquareTypes.O); RandomPlayer randomPlayer = new RandomPlayer(); NeatPlayer neatPlayer = new NeatPlayer(box, SquareTypes.X); // Play 50 games as X against a random player for (int i = 0; i < 50; i++) { // Compete the two players against each other. winner = TicTacToeGame.PlayGameToEnd(neatPlayer, randomPlayer); // Update the fitness score of the network fitness += getScore(winner, neatPlayer.SquareType); } // Play 50 games as O against a random player neatPlayer.SquareType = SquareTypes.O; for (int i = 0; i < 50; i++) { // Compete the two players against each other. winner = TicTacToeGame.PlayGameToEnd(randomPlayer, neatPlayer); // Update the fitness score of the network fitness += getScore(winner, neatPlayer.SquareType); } // Play 1 game as X against an optimal player neatPlayer.SquareType = SquareTypes.X; optimalPlayer.SquareType = SquareTypes.O; // Compete the two players against each other. winner = TicTacToeGame.PlayGameToEnd(neatPlayer, optimalPlayer); // Update the fitness score of the network fitness += getScore(winner, neatPlayer.SquareType); // Play 1 game as O against an optimal player neatPlayer.SquareType = SquareTypes.O; optimalPlayer.SquareType = SquareTypes.X; // Compete the two players against each other. winner = TicTacToeGame.PlayGameToEnd(optimalPlayer, neatPlayer); // Update the fitness score of the network fitness += getScore(winner, neatPlayer.SquareType); // Update the evaluation counter. _evalCount++; // If the network plays perfectly, it will beat the random player // and draw the optimal player. if (fitness >= 1002) { _stopConditionSatisfied = true; } // Return the fitness score return(new FitnessInfo(fitness, fitness)); }