Trait LearnedOptimizer 

pub trait LearnedOptimizer {
    // Required methods
    fn meta_train(
        &mut self,
        training_tasks: &[TrainingTask],
    ) -> Result<(), OptimizeError>;
    fn adapt_to_problem(
        &mut self,
        problem: &OptimizationProblem,
        initial_params: &ArrayView1<'_, f64>,
    ) -> Result<(), OptimizeError>;
    fn optimize<F>(
        &mut self,
        objective: F,
        initial_params: &ArrayView1<'_, f64>,
    ) -> OptimizeResult<OptimizeResults<f64>>
       where F: Fn(&ArrayView1<'_, f64>) -> f64;
    fn get_state(&self) -> &MetaOptimizerState;
    fn reset(&mut self);
}

Trait for learned optimizers

Required Methods

fn meta_train( &mut self, training_tasks: &[TrainingTask], ) -> Result<(), OptimizeError>

Meta-train the optimizer on a distribution of problems

fn adapt_to_problem( &mut self, problem: &OptimizationProblem, initial_params: &ArrayView1<'_, f64>, ) -> Result<(), OptimizeError>

Adapt to a new problem with few-shot learning

fn optimize<F>( &mut self, objective: F, initial_params: &ArrayView1<'_, f64>, ) -> OptimizeResult<OptimizeResults<f64>>

where F: Fn(&ArrayView1<'_, f64>) -> f64,

Optimize a function using learned knowledge

fn get_state(&self) -> &MetaOptimizerState

Get current meta-optimizer state

fn reset(&mut self)

Reset the optimizer

Dyn Compatibility

This trait is not dyn compatible.

In older versions of Rust, dyn compatibility was called "object safety", so this trait is not object safe.

Implementors

impl LearnedOptimizer for AdaptiveNASSystem

impl LearnedOptimizer for AdaptiveTransformerOptimizer

impl LearnedOptimizer for FewShotLearningOptimizer

impl LearnedOptimizer for LearnedHyperparameterTuner

impl LearnedOptimizer for MetaLearningOptimizer

impl LearnedOptimizer for NeuralAdaptiveOptimizer