#monte-carlo #tree-search #tree #search #carlo #monte #mcts

mocats

A fast, easy-to-use, generalized Monte Carlo Tree Search library. Works for any game, any number of players, and any tree policy (UCT Policy included as a default).

4 releases (2 breaking)

0.3.0 Oct 14, 2024
0.2.1 May 26, 2023
0.2.0 May 24, 2023
0.1.0 May 23, 2023

#782 in Algorithms

50 downloads per month

MIT license

29KB
454 lines

mocats

Crates.io License

A fast, easy-to-use, generalized Monte Carlo Tree Search library. Works for any game, any number of players, and any tree policy (UctPolicy included as a default). As of the current version, the search is single-threaded.

Features

  • Fast and efficient Monte Carlo Tree Search implementation
  • Easy-to-use API
  • Customizable number of players (uses paranoid approach for more than 2 players)
  • Customizable tree policies
  • Nicely formatted display output for debugging

Usage

In the root directory of your project, add the mocats dependency to your Cargo.toml file:

cargo add mocats

...or add this to your Cargo.toml:

[dependencies]
mocats = "0.3.0"

Defining a game

To use mocats, you must define a game and a tree policy. A game is defined by three traits:

  • GameState: Represents a game state.
  • GameAction: Represents a legal game action that can be applied to some GameState.
  • Player: Represents a player in a game. Should be an enum.

A tree policy is defined by one trait:

  • TreePolicy: Represents a tree policy.

The UctPolicy struct is included as a default tree policy.

To run the search, create a SearchTree struct with the game and tree policy, then call run on it.

use mocats::{tic_tac_toe, UctPolicy};

fn foo() {
    let game = tic_tac_toe::TicTacToePosition::new();
    let tree_policy = UctPolicy::new(2.0);
    let mut search_tree = mocats::SearchTree::new(game, tree_policy);
    search_tree.run(2000);
    let best_action = search_tree.get_best_action();
    println!("{}", search_tree);
    println!("Best action: {}", best_action.unwrap());
}

Example

See the mocats::tic_tac_toe module for a full example of implementing Tic Tac Toe using mocats. You can import tic_tac_toe to use it in your code.

use std::fmt;
use std::fmt::{Display, Formatter};

#[derive(Debug, Clone, Copy, PartialEq)]
pub struct TicTacToeMove {
    pub pos: u16
}

impl mocats::GameAction for TicTacToeMove {}

impl Display for TicTacToeMove {
    fn fmt(&self, f: &mut Formatter) -> fmt::Result {
        todo!()
    }
}

#[derive(Debug, Clone, Copy, Eq, PartialEq, Hash)]
pub enum TicTacToePlayer {
    X,
    O
}

impl TicTacToePlayer {}

impl mocats::Player for TicTacToePlayer {}

impl Display for TicTacToePlayer {
    fn fmt(&self, f: &mut Formatter<'_>) -> fmt::Result {
        todo!()
    }
}

#[derive(Debug, Clone, Copy, Eq, PartialEq, Hash)]
pub struct TicTacToePosition {
    pub board_x: u16,
    pub board_o: u16,
    pub turn: TicTacToePlayer,
}

impl TicTacToePosition {
    pub fn new() -> TicTacToePosition {
        todo!()
    }
}

impl Display for TicTacToePosition {
    fn fmt(&self, f: &mut Formatter<'_>) -> fmt::Result {
        todo!()
    }
}

impl mocats::GameState<TicTacToeMove, TicTacToePlayer> for TicTacToePosition {
    fn get_actions(&self) -> Vec<TicTacToeMove> {
        todo!()
    }

    fn apply_action(&mut self, action: &TicTacToeMove) {
        todo!()
    }

    fn get_turn(&self) -> TicTacToePlayer {
        todo!()
    }

    fn get_reward_for_player(&self, player: TicTacToePlayer) -> f32 {
        todo!()
    }
}

Documentation

For more detailed documentation and usage examples, refer to the API documentation.

Contributing

Contributions in the form of pull requests are welcome! If you encounter any issues or have suggestions for improvements, please open an issue on the GitHub repository.

License

This project is licensed under the MIT License. See the LICENSE file for details.

Dependencies

~28KB