optimizer

Bayesian and population-based optimization library with an Optuna-like API for hyperparameter tuning and black-box optimization. Supports 12 samplers, 8 pruners, multi-objective optimization, async parallelism, and persistent storage.

Quick Start

use optimizer::prelude::*;

let study: Study<f64> = Study::new(Direction::Minimize);
let x = FloatParam::new(-10.0, 10.0).name("x");

study.optimize(50, |trial| {
    let val = x.suggest(trial)?;
    Ok::<_, Error>((val - 3.0).powi(2))
}).unwrap();

let best = study.best_trial().unwrap();
println!("Best x = {:.4}, f(x) = {:.4}", best.get(&x).unwrap(), best.value);

Features at a Glance

Samplers — Random, TPE, Multivariate TPE, Grid, Sobol, CMA-ES, Gaussian Process, Differential Evolution, BOHB, NSGA-II, NSGA-III, MOEA/D
Pruners — Median, Percentile, Threshold, Patient, Hyperband, Successive Halving, Wilcoxon, Nop
Parameters — Float, Int, Categorical, Bool, and Enum types with .name() labels and typed access
Multi-objective — Pareto front extraction with NSGA-II/III and MOEA/D
Async & parallel — Concurrent trial evaluation with Tokio
Storage backends — In-memory (default) or JSONL journal for persistence and resumption
Visualization — HTML reports with optimization history and parameter importance
Analysis — fANOVA and Spearman correlation for parameter importance

Feature Flags

Flag	Enables	Default
`async`	Async/parallel optimization (Tokio)	No
`derive`	`#[derive(Categorical)]` for enum parameters	No
`serde`	Serialization of trials and parameters	No
`journal`	JSONL storage backend (implies `serde`)	No
`sobol`	Sobol quasi-random sampler	No
`cma-es`	CMA-ES sampler (requires `nalgebra`)	No
`gp`	Gaussian Process sampler (requires `nalgebra`)	No
`tracing`	Structured logging with `tracing`	No

Examples

cargo run --example basic_optimization                       # Minimize a quadratic — simplest possible usage
cargo run --example parameter_types --features derive        # All 5 param types + #[derive(Categorical)]
cargo run --example sampler_comparison                       # Compare Random, TPE, and Grid on the same problem
cargo run --example pruning                                  # Trial pruning with MedianPruner
cargo run --example early_stopping                           # Halt a study when a target is reached
cargo run --example async_parallel --features async          # Evaluate trials concurrently with tokio
cargo run --example journal_storage --features journal       # Persist trials to disk and resume later
cargo run --example ask_and_tell                             # Decouple sampling from evaluation
cargo run --example multi_objective                          # Optimize competing objectives + Pareto front

Learn More

License

MIT