- I created a ComboPlayer.java agent. It sucks and doesn't really work, but I created it. Now I'm putting it down to work on other things.

src/model/Referee.java

@@ -128,7 +128,8 @@ public class Referee implements Runnable {
 		int tile = 0;
 		for (int r = 0; r < Board.NUM_ROWS; r++) {
 			for (int c = 0; c < Board.NUM_COLS; c++) {
-				move[tile] = (mv.getCell().r == r && mv.getCell().c == c);
+				move[tile + 4] = (mv.getCell().r == r && mv.getCell().c == c);
+				tile++;
 			}
 		}
 
@@ -163,12 +164,13 @@ public class Referee implements Runnable {
 					System.out
 							.println("Interrupted while waiting for human to move!");
 				} else {
+					getPlayerModel().getOutputNodes(getBoardState(board));
+
 					Move mv = humanPlayer.getMove(board, playerModel);
 					if (board.getTile(mv.getCell().r, mv.getCell().c) == TileColor.NONE) {
 						playToken(humanPlayer.getMove(board, playerModel));
 
-						getPlayerModel().train(getBoardState(board),
+						getPlayerModel().train(getMoveArray(mv));
-								getMoveArray(mv));
 
 					} else {
 						humanPlayer.denyMove();

src/model/comPlayer/AlphaBetaComPlayer.java

@@ -4,6 +4,7 @@ import model.Board;
 import model.Move;
 import model.comPlayer.generator.AlphaBetaMoveGenerator;
 import model.comPlayer.generator.MoveGenerator;
+import model.playerModel.GameGoal;
 import model.playerModel.PlayerModel;
 
 public class AlphaBetaComPlayer implements Player {
@@ -24,6 +25,12 @@ public class AlphaBetaComPlayer implements Player {
 		return true; // always ready to play a random valid move
 	}
 
+	@Override
+	public void setGameGoal(GameGoal target) {
+		// TODO Auto-generated method stub
+
+	}
+
 	@Override
 	public String toString() {
 		return "Alpha-Beta ComPlayer";

src/model/comPlayer/ComboPlayer.java

@@ -0,0 +1,44 @@
+package model.comPlayer;
+
+import model.Board;
+import model.Move;
+import model.comPlayer.generator.AlphaBetaMoveGenerator;
+import model.comPlayer.generator.NeuralNetworkMoveGenerator;
+import model.playerModel.GameGoal;
+import model.playerModel.PlayerModel;
+
+public class ComboPlayer implements Player {
+
+	private final AlphaBetaMoveGenerator abGen = new AlphaBetaMoveGenerator();
+	private NeuralNetworkMoveGenerator nnGen = null;
+
+	@Override
+	public void denyMove() {
+		throw new UnsupportedOperationException("Not implemented");
+	}
+
+	@Override
+	public Move getMove(Board board, PlayerModel player) {
+		if (player.getHighScores()[2] == -1) {
+			return abGen.genMove(board, false);
+		}
+
+		else {
+			if (nnGen == null) {
+				nnGen = new NeuralNetworkMoveGenerator(player);
+			}
+			return nnGen.genMove(board, false);
+		}
+	}
+
+	@Override
+	public boolean isReady() {
+		// TODO Auto-generated method stub
+		return false;
+	}
+
+	@Override
+	public void setGameGoal(GameGoal target) {
+		// Nothing yet.
+	}
+}

src/model/comPlayer/HumanPlayer.java

@@ -4,6 +4,7 @@ import model.Board;
 import model.Board.TileColor;
 import model.CellPointer;
 import model.Move;
+import model.playerModel.GameGoal;
 import model.playerModel.PlayerModel;
 
 public class HumanPlayer implements Player {
@@ -79,4 +80,9 @@ public class HumanPlayer implements Player {
 	public void setColor(TileColor clr) {
 		color = clr;
 	}
+
+	@Override
+	public void setGameGoal(GameGoal target) {
+		// Do nothing.
+	}
 }

src/model/comPlayer/MinimaxComPlayer.java

@@ -4,6 +4,7 @@ import model.Board;
 import model.Move;
 import model.comPlayer.generator.MinimaxMoveGenerator;
 import model.comPlayer.generator.MoveGenerator;
+import model.playerModel.GameGoal;
 import model.playerModel.PlayerModel;
 
 public class MinimaxComPlayer implements Player {
@@ -28,6 +29,12 @@ public class MinimaxComPlayer implements Player {
 		return true; // always ready to play a random valid move
 	}
 
+	@Override
+	public void setGameGoal(GameGoal target) {
+		// TODO Auto-generated method stub
+
+	}
+
 	@Override
 	public String toString() {
 		return "Minimax ComPlayer";

src/model/comPlayer/MonteCarloComPlayer.java

@@ -4,26 +4,33 @@ import model.Board;
 import model.Move;
 import model.comPlayer.generator.MonteCarloMoveGenerator;
 import model.comPlayer.generator.MoveGenerator;
+import model.playerModel.GameGoal;
 import model.playerModel.PlayerModel;
 
 public class MonteCarloComPlayer implements Player {
-	private MoveGenerator moveGenerator = new MonteCarloMoveGenerator();
+	private final MoveGenerator moveGenerator = new MonteCarloMoveGenerator();
-
-	@Override
-	public Move getMove(Board board, PlayerModel playerModel) {
-		return moveGenerator.genMove(board, false);
-	}
 
 	@Override
 	public void denyMove() {
 		throw new UnsupportedOperationException("Not implemented");
 	}
 
+	@Override
+	public Move getMove(Board board, PlayerModel playerModel) {
+		return moveGenerator.genMove(board, false);
+	}
+
 	@Override
 	public boolean isReady() {
 		return true; // always ready to play a random valid move
 	}
 
+	@Override
+	public void setGameGoal(GameGoal target) {
+		// TODO Auto-generated method stub
+
+	}
+
 	@Override
 	public String toString() {
 		return "Monte Carlo ComPlayer";

src/model/comPlayer/NeuralNetworkPlayer.java

@@ -1,25 +1,14 @@
 package model.comPlayer;
 
 import model.Board;
-import model.Board.TileColor;
 import model.Move;
-import model.Referee;
+import model.comPlayer.generator.NeuralNetworkMoveGenerator;
-import model.playerModel.Node;
+import model.playerModel.GameGoal;
 import model.playerModel.PlayerModel;
 
 public class NeuralNetworkPlayer implements Player {
 
-	public static int getSmallest(double[] list) {
+	private NeuralNetworkMoveGenerator nnGen = null;
-		int index = 0;
-
-		for (int i = 0; i < list.length; i++) {
-			if (list[index] < list[i]) {
-				index = i;
-			}
-		}
-
-		return index;
-	}
 
 	@Override
 	public void denyMove() {
@@ -28,55 +17,11 @@ public class NeuralNetworkPlayer implements Player {
 
 	@Override
 	public Move getMove(Board board, PlayerModel player) {
-		Move mv = null;
+		if (nnGen == null) {
-
+			nnGen = new NeuralNetworkMoveGenerator(player);
-		Node[] nodes = player.getOutputNodes(Referee.getBoardState(board));
-
-		TileColor color = TileColor.BLUE;
-
-		double[] colorStrengths = new double[4];
-		colorStrengths[0] = nodes[0].strength();
-		colorStrengths[1] = nodes[1].strength();
-		colorStrengths[2] = nodes[2].strength();
-		colorStrengths[3] = nodes[3].strength();
-
-		switch (getSmallest(colorStrengths)) {
-		case 1:
-			color = TileColor.GREEN;
-			break;
-		case 2:
-			color = TileColor.RED;
-			break;
-		case 3:
-			color = TileColor.YELLOW;
-			break;
-		case 0:
-		default:
-			color = TileColor.BLUE;
 		}
 
-		int index = 4;
+		return nnGen.genMove(board, false);
-		for (int i = 4; i < nodes.length; i++) {
-			if (nodes[i].strength() > nodes[index].strength()) {
-				index = i;
-			}
-		}
-
-		int i = 4;
-		loop: for (int r = 0; r < Board.NUM_ROWS; r++) {
-			for (int c = 0; c < Board.NUM_COLS; c++) {
-				if (i == index) {
-					mv = new Move(color, r, c);
-					break loop;
-				}
-
-				else {
-					i++;
-				}
-			}
-		}
-
-		return mv;
 	}
 
 	@Override
@@ -84,6 +29,11 @@ public class NeuralNetworkPlayer implements Player {
 		return true;
 	}
 
+	@Override
+	public void setGameGoal(GameGoal target) {
+		// Do nothing.
+	}
+
 	@Override
 	public String toString() {
 		return "Neural Network Player";

src/model/comPlayer/Player.java

@@ -2,6 +2,7 @@ package model.comPlayer;
 
 import model.Board;
 import model.Move;
+import model.playerModel.GameGoal;
 import model.playerModel.PlayerModel;
 
 public interface Player {
@@ -26,4 +27,6 @@ public interface Player {
 	 * @return
 	 */
 	public boolean isReady();
+
+	public void setGameGoal(GameGoal target);
 }

src/model/comPlayer/RandomComPlayer.java

@@ -6,6 +6,7 @@ import model.Board;
 import model.Board.TileColor;
 import model.CellPointer;
 import model.Move;
+import model.playerModel.GameGoal;
 import model.playerModel.PlayerModel;
 
 public class RandomComPlayer implements Player {
@@ -51,6 +52,11 @@ public class RandomComPlayer implements Player {
 		return true; // always ready to play a random valid move
 	}
 
+	@Override
+	public void setGameGoal(GameGoal target) {
+		// TODO Auto-generated method stub
+	}
+
 	@Override
 	public String toString() {
 		return "Random ComPlayer";

src/model/comPlayer/generator/NeuralNetworkMoveGenerator.java

@@ -0,0 +1,98 @@
+package model.comPlayer.generator;
+
+import java.util.List;
+
+import model.Board;
+import model.Board.TileColor;
+import model.CellPointer;
+import model.Move;
+import model.Referee;
+import model.playerModel.Node;
+import model.playerModel.PlayerModel;
+
+public class NeuralNetworkMoveGenerator implements MoveGenerator {
+
+	public static int getSmallest(double[] list) {
+		int index = 0;
+
+		for (int i = 0; i < list.length; i++) {
+			if (list[index] < list[i]) {
+				index = i;
+			}
+		}
+
+		return index;
+	}
+
+	PlayerModel player;
+
+	public NeuralNetworkMoveGenerator(PlayerModel pm) {
+		player = pm;
+	}
+
+	@Override
+	public Move genMove(Board board, boolean asHuman) {
+		Move mv = null;
+
+		Node[] nodes = player.getOutputNodes(Referee.getBoardState(board));
+
+		TileColor color = TileColor.BLUE;
+
+		double[] colorStrengths = new double[4];
+		colorStrengths[0] = nodes[0].strength();
+		colorStrengths[1] = nodes[1].strength();
+		colorStrengths[2] = nodes[2].strength();
+		colorStrengths[3] = nodes[3].strength();
+
+		switch (getSmallest(colorStrengths)) {
+		case 1:
+			color = TileColor.GREEN;
+			break;
+		case 2:
+			color = TileColor.RED;
+			break;
+		case 3:
+			color = TileColor.YELLOW;
+			break;
+		case 0:
+		default:
+			color = TileColor.BLUE;
+		}
+
+		int index = 4;
+		for (int i = 4; i < nodes.length; i++) {
+			if (nodes[i].strength() > nodes[index].strength()) {
+				index = i;
+			}
+		}
+
+		int i = 4;
+		loop: for (int r = 0; r < Board.NUM_ROWS; r++) {
+			for (int c = 0; c < Board.NUM_COLS; c++) {
+				if (i == index) {
+					mv = new Move(color, r, c);
+					break loop;
+				}
+
+				else {
+					i++;
+				}
+			}
+		}
+
+		while (!Board.isLegal(board, mv.getCell())) {
+			mv = new Move(mv.getColor(), new CellPointer(
+					PlayerModel.rand.nextInt(Board.NUM_ROWS),
+					PlayerModel.rand.nextInt(Board.NUM_COLS)));
+		}
+
+		return mv;
+	}
+
+	@Override
+	public List<Move> genMoves(Board board, boolean asHuman, int nMoves) {
+		// Do nothing.
+		return null;
+	}
+
+}

src/model/playerModel/PlayerModel.java

@@ -134,6 +134,10 @@ public class PlayerModel implements Serializable {
 
 	public Node[] getOutputNodes(boolean[] input) {
 		if (input.length == inputNode.length) {
+			for (int i = 0; i < input.length; i++) {
+				inputNode[i].setStimulation(input[i]);
+			}
+
 			return outputNode;
 		}
 
@@ -203,12 +207,13 @@ public class PlayerModel implements Serializable {
 		}
 	}
 
-	public void train(boolean[] boardState, boolean[] example) {
+	public void train(boolean[] example) {
-		getOutputNodes(boardState);
+		boolean[] hold = getOutputActivations();
 
+		System.out.println("TRAIN");
 		if (example.length == outputNode.length) {
 			for (int i = 0; i < outputNode.length; i++) {
-				outputNode[i].learn(example[i]);
+				outputNode[i].learn(example[i] == hold[i]);
 			}
 		}
 	}
@@ -216,4 +221,14 @@ public class PlayerModel implements Serializable {
 	private int getHighScore() {
 		return getHighScores()[0];
 	}
+
+	private boolean[] getOutputActivations() {
+		boolean[] acts = new boolean[outputNode.length];
+
+		for (int i = 0; i < acts.length; i++) {
+			acts[i] = outputNode[i].axon();
+		}
+
+		return acts;
+	}
 }

src/model/playerModel/node/SigmoidNode.java

@@ -9,7 +9,7 @@ public class SigmoidNode implements Node {
 	private static final long serialVersionUID = 1L;
 
 	// Training rate.
-	private final double A = .15;
+	private final double A = .05;
 
 	private final Hashtable<Node, Double> dendrites = new Hashtable<Node, Double>();
 
@@ -36,6 +36,8 @@ public class SigmoidNode implements Node {
 				n.learn(correct);
 			}
 		}
+
+		System.out.println(strength());
 	}
 
 	@Override

src/view/ParsedArgs.java

@@ -1,17 +1,21 @@
 package view;
 
 import model.comPlayer.AlphaBetaComPlayer;
+import model.comPlayer.ComboPlayer;
 import model.comPlayer.MinimaxComPlayer;
 import model.comPlayer.MonteCarloComPlayer;
+import model.comPlayer.NeuralNetworkPlayer;
 import model.comPlayer.Player;
 import model.comPlayer.RandomComPlayer;
 
 public class ParsedArgs {
-	public static final String COM_RANDOM = "RANDOM";
-	public static final String COM_MINIMAX = "MINIMAX";
 	public static final String COM_ALPHABETA = "ALPHABETA";
-	public static final String COM_MONTECARLO = "MONTECARLO";
+	public static final String COM_ANN = "NEURALNET";
+	public static final String COM_COMBO = "COMBO";
 	public static final String COM_DEFAULT = COM_ALPHABETA;
+	public static final String COM_MINIMAX = "MINIMAX";
+	public static final String COM_MONTECARLO = "MONTECARLO";
+	public static final String COM_RANDOM = "RANDOM";
 
 	private String comPlayer = COM_DEFAULT;
 
@@ -24,8 +28,13 @@ public class ParsedArgs {
 			return new AlphaBetaComPlayer();
 		} else if (COM_MONTECARLO.equalsIgnoreCase(comPlayer)) {
 			return new MonteCarloComPlayer();
+		} else if (COM_ANN.equalsIgnoreCase(comPlayer)) {
+			return new NeuralNetworkPlayer();
+		} else if (COM_COMBO.equalsIgnoreCase(comPlayer)) {
+			return new ComboPlayer();
 		} else {
-			System.out.println("Unrecognized comPlayer '" + comPlayer +"', using default: " + COM_DEFAULT);
+			System.out.println("Unrecognized comPlayer '" + comPlayer
+					+ "', using default: " + COM_DEFAULT);
 			return new AlphaBetaComPlayer();
 		}
 	}