This repository contains some game-playing algorithms developed during my (ongoing) study of artificial intelligence.
The goal is to develop a framework containing interfaces where Games and Players implementations
can interact and experiment how different Agents algorithms fare, from classical to reinforcement
learning.
The process was split in two steps: learning and playing. That was done because in most agents the process of learning is orders of magnitude slower than the process of playing. This way it's possible to cache learned agents to avoid paying the learning cost again.
Some agents need to learn ahead of time (e.g., min-max). To do that, run
cargo run -r learn <PLAYER>. For example: cargo run -r learn min-max.
To make the agents play the games, run cargo run -r play <PLAYER_1> <PLAYER_2> <GAME_COUNT>
Example:
$ cargo run -r play min-max random 100
Win: 89.58%, Draw: 10.42%, Loss: 0.00%, Game Count: 96