Checkers Machine Learning

A checkers implementation with endpoints for machine learning.

Usage

To play checkers, run checkers.Main to start a game against a computer that makes random moves. Human input is taken from System.in in the format A2B3 to represent moving the piece on A2 to B3. To play against another human, change:

Player black = new RandomPlayer(Side.BLACK);

to:

Player black = new HumanPlayer(Side.BLACK);

Or alternatively, to watch two random computers play against each other, change the white player to a RandomPlayer.

Checkers rules are based on https://simple.wikipedia.org/wiki/Checkers.

Machine Learning

A (very simple) sample machine learning application is in package ml.example. The ExampleTrainer keeps a database of board states + moves along with a weight for each combination. After a game is completed, the results of the game are passed to the ExampleTrainer, which updates the database with the results of the game:

• all moves done by the winning player are updated with higher weights
• all moves done by the losing player are updated with lower weights

The ExampleAIPlayer queries the database based on the current state of the board and available moves, and chooses moves based on their weight (moves with higher weights have a higher chance of being chosen).

Notes

The database I am using is MySQL. The setup script is:

CREATE DATABASE checkers;
USE checkers;
CREATE TABLE states (board VARCHAR(32), move VARCHAR(4), weight FLOAT, samples INT);
ALTER TABLE states ADD UNIQUE (board, move);