Struct AdamOptimizer 

pub struct AdamOptimizer {
    pub lr: f64,
    pub beta1: f64,
    pub beta2: f64,
    pub epsilon: f64,
    pub m: Vec<f64>,
    pub v: Vec<f64>,
    pub step: u64,
}

Adam optimizer for a flat parameter vector.

Reference: Kingma & Ba (2015) “Adam: A Method for Stochastic Optimization”.

Fields

lr: f64

Learning rate α.

beta1: f64

Exponential decay rate for first moment estimates.

beta2: f64

Exponential decay rate for second moment estimates.

epsilon: f64

Small constant for numerical stability.

m: Vec<f64>

First moment vector (m).

v: Vec<f64>

Second moment vector (v).

step: u64

Current step count (t).

Implementations

impl AdamOptimizer

pub fn new( n_params: usize, lr: f64, beta1: f64, beta2: f64, epsilon: f64, ) -> Self

Create a new Adam optimizer for a parameter vector of length n_params.

pub fn default_params(n_params: usize) -> Self

Create an Adam optimizer with default hyperparameters (lr=1e-3, β1=0.9, β2=0.999, ε=1e-8).

pub fn step_update(&mut self, params: &mut [f64], grads: &[f64])

Apply one Adam update step to params using grads.

Updates params in-place and increments the step counter.

pub fn reset(&mut self)

Reset state (moments and step counter) to zero.

Trait Implementations

impl Clone for AdamOptimizer

fn clone(&self) -> AdamOptimizer

Returns a duplicate of the value.

fn clone_from(&mut self, source: &Self)

Performs copy-assignment from source.

impl Debug for AdamOptimizer

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.