Connect4 implementation for Alpha Zero General

Alpha-zero general implementation of connect4. Neural network architecture was copy-pasted from the game of Othello, so could likely be improved.

To train a model for Connect4, update the imports in main.py to:

from Coach import Coach
from connect4.Connect4Game import Connect4Game
from connect4.tensorflow.NNet import NNetWrapper as nn
from utils import dotdict

and the first line of __main__ to

g = Connect4Game()

Make similar changes to pit.py.

To start training a model for Connect4:

python main.py

To start a tournament of 100 episodes with the model-based player against a random player:

python pit.py

You can play againt the model by switching to HumanPlayer in pit.py

Experiments

Experiments so far have been CPU constrained. I trained a tensorflow model for 6x7 Connect4 board (100 episodes, 10 epochs per iteration and 25 MCTS simulations per turn) for 6 iterations which took about ~5 hours on an old i7-2670QM laptop without CUDA. Running a further 6 iterations (20 hours?) has not yielded a noticable improvment.

After 6 iterations, the model is wins 50% of the time vs the 'OneStepLookaheadConnect4Player' player.

Experiments 2

I trained a tensorflow model for 6x7 Connect4 board (100 episodes, 10 epochs per iteration and 25 MCTS simulations per turn) for 32 iterations which took about ~5 hours on a i7-6700HQ laptop

After 32 iterations, the model get a pretty good result (83 win,7 lose,10 draw) vs the 'OneStepLookaheadConnect4Player' player.