Struct connect6::policy::AlphaZero

pub struct AlphaZero { /* fields omitted */ }

Implementation of policy AlphaZero based on combined MCTS with non-linear value approximator.

AlphaZero policy is implemented based on Mastering the game of Go with deep neural networks and tree search and Mastering Chess and Shogi by Self-Play with a General Reinforcement Learning Algorithm.

Examples

let param = HyperParameter::light_weight();
let mut policy = AlphaZero::with_param(Box::new(RandomEvaluator {}), param);
let result = Agent::new(&mut policy).play();
assert!(result.is_ok());

Methods

impl AlphaZero

pub fn new(evaluator: Box<dyn Evaluator + Send>) -> AlphaZero

Construct a new AlphaZero policy with given evalueator

pub fn with_param(
    evaluator: Box<dyn Evaluator + Send>,
    param: HyperParameter
) -> AlphaZero

Construct a AlphaZero with given hyperparam

Trait Implementations

impl Policy for AlphaZero

fn next(&mut self, game: &Game) -> Option<(usize, usize)>

Select next position with AlphaZero policy