THIS FOLDER CONTAINS THE IMPLEMENTATION OF THE Q-LEARNING ALGORITHM

TO PLAY TIC TAC TOE

COPYRIGHT BELONGS TO THE AUTHOR OF THIS REPOSITORY

AUTHOR : LAKSHMAN KUMAR

AFFILIATION : UNIVERSITY OF MARYLAND, MARYLAND ROBOTICS CENTER
EMAIL : LKUMAR93@UMD.EDU
LINKEDIN : WWW.LINKEDIN.COM/IN/LAKSHMANKUMAR1993

THE WORK (AS DEFINED BELOW) IS PROVIDED UNDER THE TERMS OF THE MIT LICENSE
THE WORK IS PROTECTED BY COPYRIGHT AND/OR OTHER APPLICABLE LAW. ANY USE OF
THE WORK OTHER THAN AS AUTHORIZED UNDER THIS LICENSE OR COPYRIGHT LAW IS
PROHIBITED.

BY EXERCISING ANY RIGHTS TO THE WORK PROVIDED HERE, YOU ACCEPT AND AGREE TO
BE BOUND BY THE TERMS OF THIS LICENSE. THE LICENSOR GRANTS YOU THE RIGHTS
CONTAINED HERE IN CONSIDERATION OF YOUR ACCEPTANCE OF SUCH TERMS AND
CONDITIONS.

INSTRUCTIONS

- Ubuntu 14.04 and Python 2.7+ are required to use this repository

- Install the required libraries using the following commands

1) sudo apt-get install python-pygame
2) sudo apt-get install python-matplotlib
3) sudo apt-get install python-pip
4) pip install h5py
5) sudo apt-get install libhdf5-serial-dev
6) sudo apt-get install python-numpy python-scipy python-dev python-pip
python-nose
g++ libopenblas-dev git liblapack-dev gfortran
7) sudo pip install Theano
8) sudo pip install keras

- Once you install keras, change the backend to theano as shown in the link below
https://keras.io/backend/

- The folder contains 4 source code files, TicTacToe.py contains the

implementation of the
TicTacToe game , QLearning.py contains the implementation of the Q Learning
Algorithm,
Train.py is used to train the Qlearner for 30,000 episodes by self play and
Play.py can be
used by the user to play with the Qlearner. The learnt policies by the qlearner
have been
saved as either a JSON file or a HDF5 file based on whether a look up table is
used or
function approximation is used

- Usage :

Training -

python Train.py -f <bool> -r <bool> -e <bool>

By default, the qlearner trains itself by playing against another

qlearner
 using Look up table

To enable function approximation, experience replay and a random

opponent ,
use the following command

python Train.py -f True -r True -r True

Testing -

python Play.py -f <bool> -e <bool>

By default, the qlearner plays against the user using Look Up Table

To enable function approximation and experience replay ,

use the following command

python Play.py -f True -e True