Using Generic Programming to evolve a CONNECT FOUR playing AI

Requirements

• Java 7
• An unix-shell
• An Ant build system

Building

Releases can be found online.

Using Ant

ant compile

Using Eclipse

Help can you find here.

Running

From your console:

java -jar gp_connect_four.jar

Notes on running

You can set the options in the Main.java file.

In the main method you can select what you want to do:

• Play against a player (calling the play method), or
• let the evolve a player for you (calling the evolve method).

You then have to compile the program, as of now no command line options have been integrated. Can take a long while with the right/wrong options.

Saving an evolved player

1. Run the programing evolving an player. Save the code yourself, or redirect its output like this java -jar gp_connect_four.jar > mi_01_only_ternary.txt

2. Create a new class in the package org.samuelsmal.game.EvolvedPlayers, extending GPPlayer.

3. Implement the required methods (see below).

4. Insert into the play()-method of the Main-class in the Main.java-file an new Instance of your new class created in step 2, replace one of the old players. The argument sequence corresponds to the start player.

   @Override protected long evalGame(char playerColour, char enemyColour, Game game) { // PAST CODE FROM PROGRAMM HERE }

Using the data

You can extract the data using grep:

grep '^[0-9]' mi_01_only_ternary.txt > mi_01_only_ternary_points.txt

This will extract all the datapoints. If you want to extract only the winner use

grep '^0' mi_01_only_ternary.txt > mi_01_only_ternary_only_winner.txt

If you want to extract the number of nodes use:

grep 'Winner count of nodes' mi_01_only_ternary.txt > mi_01_only_ternary_node_count.txt

And some replacement tool of your own.

You can use my Gist to plot the data, it's written in R.