Struct WeightedOptimizer 

pub struct WeightedOptimizer<A, D>
where
    A: Float + ScalarOperand + Debug,
    D: Dimension,
{ /* private fields */ }

A weighted composition of optimizers

Runs all optimizers on the same parameters/gradients and returns the weighted average of their outputs. This allows blending the behavior of multiple optimization strategies.

Example

use scirs2_core::ndarray::Array1;
use optirs_core::optimizer_composition::WeightedOptimizer;
use optirs_core::optimizers::{SGD, Adam, Optimizer};

// Create a weighted combination of SGD and Adam
let mut weighted = WeightedOptimizer::new()
    .add_optimizer(Box::new(SGD::new(0.1)), 0.7)
    .add_optimizer(Box::new(Adam::new(0.01)), 0.3);

let params = Array1::zeros(3);
let gradients = Array1::ones(3);
let updated = weighted.step(&params, &gradients).expect("step failed");

Implementations

impl<A, D> WeightedOptimizer<A, D>

where A: Float + ScalarOperand + Debug, D: Dimension,

pub fn new() -> Self

Create a new empty weighted optimizer

pub fn add_optimizer(self, opt: Box<dyn Optimizer<A, D>>, weight: A) -> Self

Add an optimizer with a given weight (builder pattern)

Arguments

• opt - The optimizer to add
• weight - The weight for this optimizer

pub fn with_optimizers(self, opts: Vec<(Box<dyn Optimizer<A, D>>, A)>) -> Self

Add multiple optimizers at once (builder pattern)

Arguments

• opts - A vector of (optimizer, weight) pairs

pub fn normalize_weights(&mut self)

Normalize weights so they sum to 1

pub fn num_optimizers(&self) -> usize

Get the number of optimizers

pub fn weights(&self) -> &[A]

Get the current weights

Trait Implementations

impl<A, D> Default for WeightedOptimizer<A, D>

where A: Float + ScalarOperand + Debug, D: Dimension,

fn default() -> Self

Returns the “default value” for a type.

impl<A, D> Optimizer<A, D> for WeightedOptimizer<A, D>

where A: Float + ScalarOperand + Debug, D: Dimension,

fn step( &mut self, params: &Array<A, D>, gradients: &Array<A, D>, ) -> Result<Array<A, D>>

Updates parameters using the given gradients

fn get_learning_rate(&self) -> A

Gets the current learning rate

fn set_learning_rate(&mut self, learning_rate: A)

Sets a new learning rate

fn step_list( &mut self, params_list: &[&Array<A, D>], gradients_list: &[&Array<A, D>], ) -> Result<Vec<Array<A, D>>>

Updates multiple parameter arrays at once