Trait SimdOptimizer 

pub trait SimdOptimizer<T: Float> {
    // Required methods
    fn simd_sgd_update(
        params: &ArrayView1<'_, T>,
        gradients: &ArrayView1<'_, T>,
        learning_rate: T,
    ) -> Array1<T>;
    fn simd_momentum_update(
        params: &ArrayView1<'_, T>,
        gradients: &ArrayView1<'_, T>,
        velocity: &ArrayView1<'_, T>,
        learning_rate: T,
        momentum: T,
    ) -> (Array1<T>, Array1<T>);
    fn simd_adam_first_moment(
        m: &ArrayView1<'_, T>,
        gradients: &ArrayView1<'_, T>,
        beta1: T,
    ) -> Array1<T>;
    fn simd_adam_second_moment(
        v: &ArrayView1<'_, T>,
        gradients: &ArrayView1<'_, T>,
        beta2: T,
    ) -> Array1<T>;
    fn simd_adam_update(
        params: &ArrayView1<'_, T>,
        m_hat: &ArrayView1<'_, T>,
        v_hat: &ArrayView1<'_, T>,
        learning_rate: T,
        epsilon: T,
    ) -> Array1<T>;
    fn simd_weight_decay(
        gradients: &ArrayView1<'_, T>,
        params: &ArrayView1<'_, T>,
        weight_decay: T,
    ) -> Array1<T>;
    fn simd_gradient_norm(gradients: &ArrayView1<'_, T>) -> T;
}

Trait for SIMD-accelerated optimizer operations

This trait provides high-performance implementations of common operations found in optimization algorithms.

Required Methods

fn simd_sgd_update( params: &ArrayView1<'_, T>, gradients: &ArrayView1<'_, T>, learning_rate: T, ) -> Array1<T>

SIMD-accelerated parameter update: params - learning_rate * gradient

Arguments

• params - Parameter array
• gradients - Gradient array
• learning_rate - Learning rate scalar

Returns

Updated parameters

fn simd_momentum_update( params: &ArrayView1<'_, T>, gradients: &ArrayView1<'_, T>, velocity: &ArrayView1<'_, T>, learning_rate: T, momentum: T, ) -> (Array1<T>, Array1<T>)

SIMD-accelerated momentum update

velocity = momentum * velocity + learning_rate * gradient params = params - velocity

Arguments

• params - Parameter array
• gradients - Gradient array
• velocity - Velocity array (momentum state)
• learning_rate - Learning rate scalar
• momentum - Momentum coefficient

Returns

Tuple of (updated_params, updated_velocity)

fn simd_adam_first_moment( m: &ArrayView1<'_, T>, gradients: &ArrayView1<'_, T>, beta1: T, ) -> Array1<T>

SIMD-accelerated Adam first moment update

m = beta1 * m + (1 - beta1) * gradient

Arguments

• m - First moment array
• gradients - Gradient array
• beta1 - Exponential decay rate for first moment

Returns

Updated first moment

fn simd_adam_second_moment( v: &ArrayView1<'_, T>, gradients: &ArrayView1<'_, T>, beta2: T, ) -> Array1<T>

SIMD-accelerated Adam second moment update

v = beta2 * v + (1 - beta2) * gradient^2

Arguments

• v - Second moment array
• gradients - Gradient array
• beta2 - Exponential decay rate for second moment

Returns

Updated second moment

fn simd_adam_update( params: &ArrayView1<'_, T>, m_hat: &ArrayView1<'_, T>, v_hat: &ArrayView1<'_, T>, learning_rate: T, epsilon: T, ) -> Array1<T>

SIMD-accelerated Adam parameter update

params = params - learning_rate * m_hat / (sqrt(v_hat) + epsilon)

Arguments

• params - Parameter array
• m_hat - Bias-corrected first moment
• v_hat - Bias-corrected second moment
• learning_rate - Learning rate scalar
• epsilon - Small constant for numerical stability

Returns

Updated parameters

fn simd_weight_decay( gradients: &ArrayView1<'_, T>, params: &ArrayView1<'_, T>, weight_decay: T, ) -> Array1<T>

SIMD-accelerated weight decay application

gradients = gradients + weight_decay * params

Arguments

• gradients - Gradient array
• params - Parameter array
• weight_decay - Weight decay coefficient

Returns

Gradients with weight decay applied

fn simd_gradient_norm(gradients: &ArrayView1<'_, T>) -> T

SIMD-accelerated gradient norm computation

Arguments

• gradients - Gradient array

Returns

L2 norm of gradients

Dyn Compatibility

This trait is not dyn compatible.

In older versions of Rust, dyn compatibility was called "object safety".

Implementations on Foreign Types

impl SimdOptimizer<f32> for f32

Implementation of SIMD optimizer operations for f32

fn simd_sgd_update( params: &ArrayView1<'_, f32>, gradients: &ArrayView1<'_, f32>, learning_rate: f32, ) -> Array1<f32>

fn simd_momentum_update( params: &ArrayView1<'_, f32>, gradients: &ArrayView1<'_, f32>, velocity: &ArrayView1<'_, f32>, learning_rate: f32, momentum: f32, ) -> (Array1<f32>, Array1<f32>)

fn simd_adam_first_moment( m: &ArrayView1<'_, f32>, gradients: &ArrayView1<'_, f32>, beta1: f32, ) -> Array1<f32>

fn simd_adam_second_moment( v: &ArrayView1<'_, f32>, gradients: &ArrayView1<'_, f32>, beta2: f32, ) -> Array1<f32>

fn simd_adam_update( params: &ArrayView1<'_, f32>, m_hat: &ArrayView1<'_, f32>, v_hat: &ArrayView1<'_, f32>, learning_rate: f32, epsilon: f32, ) -> Array1<f32>

fn simd_weight_decay( gradients: &ArrayView1<'_, f32>, params: &ArrayView1<'_, f32>, weight_decay: f32, ) -> Array1<f32>

fn simd_gradient_norm(gradients: &ArrayView1<'_, f32>) -> f32

impl SimdOptimizer<f64> for f64

Implementation of SIMD optimizer operations for f64

fn simd_sgd_update( params: &ArrayView1<'_, f64>, gradients: &ArrayView1<'_, f64>, learning_rate: f64, ) -> Array1<f64>

fn simd_momentum_update( params: &ArrayView1<'_, f64>, gradients: &ArrayView1<'_, f64>, velocity: &ArrayView1<'_, f64>, learning_rate: f64, momentum: f64, ) -> (Array1<f64>, Array1<f64>)

fn simd_adam_first_moment( m: &ArrayView1<'_, f64>, gradients: &ArrayView1<'_, f64>, beta1: f64, ) -> Array1<f64>

fn simd_adam_second_moment( v: &ArrayView1<'_, f64>, gradients: &ArrayView1<'_, f64>, beta2: f64, ) -> Array1<f64>

fn simd_adam_update( params: &ArrayView1<'_, f64>, m_hat: &ArrayView1<'_, f64>, v_hat: &ArrayView1<'_, f64>, learning_rate: f64, epsilon: f64, ) -> Array1<f64>

fn simd_weight_decay( gradients: &ArrayView1<'_, f64>, params: &ArrayView1<'_, f64>, weight_decay: f64, ) -> Array1<f64>

fn simd_gradient_norm(gradients: &ArrayView1<'_, f64>) -> f64

Implementors