Noughts and Crosses
As a way to learn Tensorflow and Deep Neural Networks, I made a DNN model to play Noughts and Crosses. It’s interesting from a technical point of view, but ultimately is not the best way to solve the game… search would be a better algorithm, and if you do want to use DNNs, ideally you would use Reinforcement Learning techniques too. The game play model is probably the most reusable part.
A full writeup is here: Noughts and Crosses