It is required to developed artificial intelligence based on neural networks for backgammon. Artificial intelligence has to play better than TD-Gammon (http://en.wikipedia.org/wiki/TD-Gammon).
We already have the source code with game logic, it is only required to add the neural network and force it to learn. The neural network should be saved in files (for example, sqlite).
We expect to have separate files for different difficulty levels. The more games are trained via neural network, the more intelligent artificial intellect becomes. Neural network has to be trained until it will have played at least 2 million games.
The game quality cannot be lower than that in TD-Gammon. Here you can read more about neural network realization process in backgammon:
This code might help the task.
Programming language used for realization - C++.
HOW GAME LOGIC...