TTT Agent

This is an implementation of a multiagent value iteration algorithm for playing Tic Tac Toe.

Primer

The problem can be formulated into a Markov Decision Process (MDP) in which we assess elements such as environment, state, action, and value. The implementation leverages a dynamic programming algorithm called value iteration to find the optimal action for any state in the tic-tac-toe environment.

Gameplay

The following are guidelines on how to train reinforcement learning agents as well as how you can play your trained agents.

Building

The program executable can be built by using the cmake build tool.

$ cmake .
$ make
$ ./ttt-agent

Training

You will be able to specify the experiment rate as well as the number of games to simulate during training. The experiment rate is used to randomly select an action during the training process in order to alleviate overfitting issues.

$ ./ttt-agent
Would you like to train new TTT agents(y,n) y
Input experiment rate: 0.3
How many rounds would you like to simulate: 10000
-------
| |O|X|
-------
| |X| |
-------
|X| |O|
-------
55% [=====>     ] 5510 rnds
Duration: 0m37s
148.92 Rnds/sec

Human Player

There is also an option to play against the recently trained agent.

$ ./ttt-agent
Would you like to train new TTT agents(y,n) n
What symbol would you like to play as(X,O) O
-------
| | | | 1
-------
| | | | 2
-------
| | |X| 3
-------
 1 2 3
Row: 2
Column: 2
-------
| | | | 1
-------
| |O| | 2
-------
| |X|X| 3
-------
 1 2 3
Row: