Struct AdamOptimizer 

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

Adam adaptive moment estimator.

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

Fields

beta1: f64

First moment decay rate (typically 0.9).

beta2: f64

Second moment decay rate (typically 0.999).

eps: f64

Numerical stability constant (typically 1e-8).

lr: f64

Learning rate.

m: Vec<f64>

First moment (mean) buffer.

v: Vec<f64>

Second moment (variance) buffer.

t: u64

Current time-step (number of update calls so far).

Implementations

impl AdamOptimizer

pub fn new(n: usize, lr: f64, beta1: f64, beta2: f64, eps: f64) -> Self

Create a new Adam optimiser for n parameters.

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

Apply one Adam update step.

• params — mutable slice of parameter values.
• grads — gradient slice of the same length.

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.