Deveolop a machine learning algorithm based off deepmind's AlphaZero that is able to learn how to play two player games of perfect information (Ex: chess, checkers, dots and boxes, coridor ect..) simply by playing itself. While the ai will have some specific inputs/outputs based on the game it should be able to learn and hopefully beat some humans without any additonal human data.
Some notable components:
- Neural Networks
- Monte Carlo Search Trees
Implementation details:
- In python that way I can easily integrate it with ML libraries like tensor flow but I might try to make my own ML lib from scratch.
- Start off with a desktop application with possibiliy of putting a trained model on a website
Article: https://deepmind.com/blog/article/alphazero-shedding-new-light-grand-games-chess-shogi-and-go