Struct AdaGradOptimizer 

pub struct AdaGradOptimizer {
    pub lr: f64,
    pub eps: f64,
    pub accum: Vec<f64>,
}

AdaGrad optimizer.

Adapts the learning rate for each parameter by accumulating squared gradients. Parameters that receive large, frequent gradients see smaller effective learning rates.

Update rule

G_t = G_{t-1} + g_t ⊙ g_t
θ_t = θ_{t-1} - α / (√G_t + ε) ⊙ g_t

Reference: Duchi et al. (2011).

Fields

lr: f64

Global learning rate

eps: f64

Numerical stability constant

accum: Vec<f64>

Accumulated squared gradients

Implementations

impl AdaGradOptimizer

pub fn new(lr: f64, eps: f64) -> Self

Create a new AdaGrad optimizer.

pub fn default_params(lr: f64) -> Self

Create with default hyperparameters (lr=0.01, eps=1e-8).

pub fn step( &mut self, params: &mut Vec<f64>, grad: &[f64], ) -> Result<(), OptimizeError>

Perform one AdaGrad update step.

Errors

Returns OptimizeError::ValueError if length mismatch.

pub fn reset(&mut self)

Reset accumulated state.

Trait Implementations

impl Clone for AdaGradOptimizer

fn clone(&self) -> AdaGradOptimizer

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for AdaGradOptimizer

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

Formats the value using the given formatter.