OptunaSharp 1.0.0

OptunaSharp

Optuna for .NET: a C#/F# port of Optuna, the define-by-run hyperparameter optimization framework, built on NumSharp.

Install

dotnet add package OptunaSharp

The package targets .NET 8 and .NET 10.

Quick start

using OptunaSharp;

var study = Optuna.CreateStudy(direction: StudyDirection.Minimize);

study.Optimize(trial =>
{
    double x = trial.SuggestFloat("x", -10, 10);
    return (x - 2) * (x - 2);
}, nTrials: 100);

Console.WriteLine(study.BestValue);        // ~0
Console.WriteLine(study.BestParams["x"]);  // ~2

What this package contains

• Studies and trials: Optuna.CreateStudy / LoadStudy, Optimize, ask-and-tell, multi-objective directions, constraints, user and system attributes, callbacks.
• Distributions and search spaces: float, int and categorical distributions, plus search-space inference for dynamic (define-by-run) spaces.
• Samplers: TPE, CMA-ES, GP, Random, Grid, QMC, BruteForce, PartialFixed, NSGA-II and NSGA-III. It also ships samplers Optuna does not have, among them AGE-MOEA-2, SMS-EMOA, MOEA/D, RVEA, LM-CMA, VkD-CMA, differential evolution, PSO, the natural evolution strategies (OpenAI-ES, SNES, xNES), TuRBO, REMBO, SAASBO and JANUS.
• Pruners: Median, Percentile, SuccessiveHalving, Hyperband, Threshold, Patient, Wilcoxon and Nop.
• Storages: in-memory and journal-file storage, heartbeats and RetryFailedTrialCallback.
• Analysis: parameter importance (fANOVA, mean decrease impurity, PED-ANOVA, SHAP), terminators, the data layer behind every plot_* function, and a scikit-learn-style OptunaSearchCV.
• Artifacts: a file-system artifact store.

Companion packages

Anything that needs a foreign runtime or a heavy client library ships separately, so this package depends only on NumSharp, its OpenBLAS backend and Microsoft.Extensions.DependencyInjection.Abstractions.

Compatibility target

OptunaSharp tracks Optuna 5.x as its behavioral source of truth: names, defaults, edge cases and error messages follow Optuna. Where OptunaSharp and Optuna differ, Optuna is treated as correct and OptunaSharp is aligned to it. The test suite checks OptunaSharp against live Python Optuna, including seeded runs that must reproduce the values Optuna samples exactly.

Links

• Documentation: https://scisharp.github.io/OptunaSharp
• Source and issues: https://github.com/SciSharp/OptunaSharp

License

MIT. OptunaSharp is a derivative of Optuna (© Preferred Networks, Inc.) and retains its notice — see LICENSE.