Struct HyperparameterOptimizer 

pub struct HyperparameterOptimizer { /* private fields */ }

Gradient-based hyperparameter optimizer

Implementations

impl HyperparameterOptimizer

pub fn new(config: HyperparameterConfig) -> Self

Create a new hyperparameter optimizer

pub fn add_hyperparameter(&mut self, hyperparameter: OptimizableHyperparameter)

Add a hyperparameter to optimize

pub fn get_hyperparameter(&self, name: &str) -> Result<f64>

Get hyperparameter value by name

pub fn step<F>(&mut self, objective_fn: F) -> Result<f64>

where F: Fn(&HashMap<String, f64>) -> Result<Tensor>,

Perform one step of hyperparameter optimization

pub fn optimize<F, V>( &mut self, objective_fn: F, validation_fn: V, ) -> Result<HyperparameterOptimizationResult>

where F: Fn(&HashMap<String, f64>) -> Result<Tensor>, V: Fn(&HashMap<String, f64>) -> Result<f64>,

Optimize hyperparameters using validation loss

impl HyperparameterOptimizer

Convenience functions for common hyperparameter optimization scenarios

pub fn for_learning_rate( initial_lr: f64, config: Option<HyperparameterConfig>, ) -> Result<Self>

Create optimizer for learning rate optimization

pub fn for_regularization( initial_reg: f64, config: Option<HyperparameterConfig>, ) -> Result<Self>

Create optimizer for regularization strength

pub fn for_multiple_params( params: Vec<(&str, f64, Option<f64>, Option<f64>, bool)>, config: Option<HyperparameterConfig>, ) -> Result<Self>

Create optimizer for multiple hyperparameters