From number of iterations, execute episodes and othello arena in parallel.

Add Othello paper author's CNN weights to our NN structure

Create class to represent the game

Change the neural network architecture used on Agent to a ResNet with shared features, the same strategy used on AlphaGo Zero.

create a new implementation of the neural network, using TensorFlow or PyTorch, to obtain a better performance during training.

Create Python script to load saved weights and enable human to play with the treined model. (For now lets just play using the terminal)

From NN object and number of MCTS simulations, execute an episode

From two NN objects play a game using these two models and return the NN object winner.

Class to create NN structure with methods to train it