Module stochastic

Stochastic optimization methods for machine learning and large-scale problems

This module provides stochastic optimization algorithms that are particularly well-suited for machine learning, neural networks, and large-scale problems where exact gradients are expensive or noisy.

Re-exports

pub use adam::minimize_adam;

pub use adam::AdamOptions;

pub use adamw::minimize_adamw;

pub use adamw::AdamWOptions;

pub use momentum::minimize_sgd_momentum;

pub use momentum::MomentumOptions;

pub use rmsprop::minimize_rmsprop;

pub use rmsprop::RMSPropOptions;

pub use sgd::minimize_sgd;

pub use sgd::SGDOptions;

Modules

adam

ADAM (Adaptive Moment Estimation) optimizer

adamw

AdamW (Adam with decoupled Weight Decay) optimizer

momentum

SGD with Momentum optimizer

rmsprop

RMSProp (Root Mean Square Propagation) optimizer

sgd

Stochastic Gradient Descent (SGD) optimization

Structs

BatchGradientWrapper

Wrapper for regular gradient functions

InMemoryDataProvider

Simple in-memory data provider

StochasticOptions

Common options for stochastic optimization

Enums

LearningRateSchedule

Learning rate schedules

StochasticMethod

Stochastic optimization method selection

Traits

DataProvider

Data provider trait for stochastic optimization

StochasticGradientFunction

Stochastic gradient function trait

Functions

clip_gradients

Clip gradients to prevent exploding gradients

create_stochastic_options_for_problem

Create stochastic options optimized for specific problem types

generate_batch_indices

Generate random batch indices

minimize_stochastic

Main stochastic optimization function

update_learning_rate

Update learning rate according to schedule