Table of Contents

Quick visualization

OptunaSharp ports Optuna's visualization suite so you can analyze a study visually — the optimization history, learning curves, parameter relationships, and importances. Install OptunaSharp.Visualization.Plotly for ready-made interactive charts (PlotlyVisualization), or consume the Core data-layer builders and render however you like.

A study to visualize

using OptunaSharp;
using OptunaSharp.Samplers;
using OptunaSharp.Pruners;
using OptunaSharp.Visualization.Plotly;   // OptunaSharp.Visualization.Plotly package

const int seed = 13;
var study = Optuna.CreateStudy(
    direction: StudyDirection.Maximize,
    sampler: new TPESampler(seed: seed),
    pruner: new MedianPruner());
study.Optimize(Objective, nTrials: 30, timeout: 300);

Plot functions

Each returns a Plotly.NET GenericChart you can display or export:

// Objective value and best-so-far over trials.
var history   = PlotlyVisualization.PlotOptimizationHistory(study);

// The reported intermediate values (learning curves) per trial.
var learning  = PlotlyVisualization.PlotIntermediateValues(study);

// All parameters and the objective as parallel axes…
var parallel  = PlotlyVisualization.PlotParallelCoordinate(study);
// …or a chosen subset.
var parallel2 = PlotlyVisualization.PlotParallelCoordinate(study, params: new[] { "lr", "n_layers" });

// Objective over parameter pairs.
var contour   = PlotlyVisualization.PlotContour(study, params: new[] { "lr", "n_layers" });

// Objective vs. each parameter.
var slice     = PlotlyVisualization.PlotSlice(study, params: new[] { "lr", "n_layers" });

// Parameter importances as bars.
var importance = PlotlyVisualization.PlotParamImportances(study);

// Importance with respect to a custom target — e.g. which params drive trial duration.
var durImp    = PlotlyVisualization.PlotParamImportances(
    study, target: t => t.Duration!.Value.TotalSeconds, targetName: "duration");

// Empirical distribution of the objective values.
var edf       = PlotlyVisualization.PlotEdf(study);

// Trials ranked over parameter pairs, colored by objective.
var rank      = PlotlyVisualization.PlotRank(study);

// Start/finish timeline of the trials.
var timeline  = PlotlyVisualization.PlotTimeline(study);

The Visualization guide lists every builder and the info record it returns; the Parameter Importance guide covers the importance evaluators behind PlotParamImportances.

Customizing a figure

Each plot returns an editable Plotly.NET chart, so you can adjust titles, axes, and layout with the Plotly.NET API — or pass a PlotlyChartCustomization to the customization overload:

var fig = PlotlyVisualization.PlotIntermediateValues(study)
    .WithTitle("Hyperparameter optimization for FashionMNIST classification");
  • Multi-objective: the Pareto-front plot is covered in the multi-objective recipe (PlotParetoFront).
  • Optuna Dashboard: persist the study to an RDB backend and the Python optuna-dashboard tool renders the same history, importances, and relationships in a live web UI — it reads OptunaSharp's SQLite files directly.

Relationship to Optuna

Optuna (Python) OptunaSharp
optuna.visualization.plot_optimization_history(study) PlotlyVisualization.PlotOptimizationHistory(study)
plot_contour(study, params=["lr", "n_layers"]) PlotlyVisualization.PlotContour(study, params: new[] { "lr", "n_layers" })
plot_param_importances(study, target=lambda t: t.duration.total_seconds(), target_name="duration") PlotlyVisualization.PlotParamImportances(study, target: t => t.Duration!.Value.TotalSeconds, targetName: "duration")
plot_slice / plot_edf / plot_rank / plot_timeline PlotSlice / PlotEdf / PlotRank / PlotTimeline

See also

Visualization guide · Parameter Importance · Plotly.NET integration · Multi-objective recipe · the API reference.

Edit this page