Struct ExperienceBuffer 

pub struct ExperienceBuffer {
    pub states: Vec<GameState>,
    pub actions: Vec<Action>,
    pub rewards: Vec<f32>,
    pub capacity: usize,
}

Buffer for storing experience tuples for learning.

Fields

states: Vec<GameState>

actions: Vec<Action>

rewards: Vec<f32>

capacity: usize

Implementations

impl ExperienceBuffer

pub fn new(capacity: usize) -> Self

pub fn push(&mut self, state: GameState, action: Action, reward: f32)

pub fn len(&self) -> usize

pub fn is_empty(&self) -> bool

pub fn sample_indices(&self, batch_size: usize, rng_seed: u64) -> Vec<usize>

Sample a random mini-batch of indices.

pub fn compute_returns(&self, gamma: f32) -> Vec<f32>

Compute discounted returns from the reward sequence.

pub fn clear(&mut self)

pub fn mean_reward(&self) -> f32

Mean reward across all stored experiences.