Monte Carlo Model Tree for Connect 4

[caleb-vicente]

Check this implementation: https://medium.com/towards-data-science/deep-reinforcement-learning-and-monte-carlo-tree-search-with-connect-4-ba22a4713e7a

Create the following:

• [ ] Base class for MTCS
• [ ] Create random simulation
• [ ] Create evironment for tip tap toe
• [ ] Run random agent vs MTCS to check that MTCS is working for tip tap toe
• [ ] Create NN simulation
• [ ] Create environment for Connect 4
• [ ] Run random agent vs MTCS to check that MTCS is working for connect 4

[caleb-vicente]

The main issue with the previous implementations is that it does´t have any type of abstration such as a neural network. It is only saving states and learning from them. Alpha Zero, on the contrary, makes the rollout stage based on a self-learning nn. https://web.stanford.edu/~surag/posts/alphazero.html