Module multi_fidelity 

Multi-fidelity optimization: Hyperband and Successive Halving.

This module provides resource-efficient hyperparameter optimization via early-stopping strategies.

• SuccessiveHalving allocates a fixed budget across configurations and iteratively discards the worst performers.
• Hyperband runs multiple brackets of Successive Halving with varying aggressiveness to hedge against the unknown ideal trade-off.

Quick start

use scirs2_optimize::multi_fidelity::{
    Hyperband, MultiFidelityConfig, ConfigSampler,
};

let config = MultiFidelityConfig {
    max_budget: 81.0,
    min_budget: 1.0,
    eta: 3,
    ..Default::default()
};

let hb = Hyperband::new(config).expect("valid config");
let bounds = vec![(-5.0, 5.0), (-5.0, 5.0)];
let mut rng = 42u64;

let result = hb.run(
    &|x: &[f64], _budget: f64| Ok(x.iter().map(|v| v * v).sum()),
    &bounds,
    &ConfigSampler::Random,
    &mut rng,
).expect("optimization ok");

println!("Best: {:?} -> {:.6}", result.best_config, result.best_objective);

Re-exports

pub use hyperband::Hyperband;

pub use successive_halving::SuccessiveHalving;

pub use types::ConfigSampler;

pub use types::EvaluationResult;

pub use types::MultiFidelityConfig;

pub use types::MultiFidelityResult;

Modules

hyperband

Hyperband algorithm.

successive_halving

Successive Halving (SHA) algorithm.

types

Types for multi-fidelity optimization (Hyperband / Successive Halving).