Visualization

OptunaSharp ports Optuna's visualization suite as a data layer: each plot is a builder that computes the exact numbers Optuna would hand to Plotly — axes, series, colors, hover text — returned as a strongly-typed info record. You can render with the implemented Plotly.NET integration, or consume these records in your own charting library. This split keeps Core free of a plotting dependency while still guaranteeing the figures are computed identically to Optuna. Everything lives in OptunaSharp.Visualization.

The builders below return info records (the _get_*_info analog). Install OptunaSharp.Visualization.Plotly for ready-made interactive figures; its integration guide covers chart methods and HTML output.

Using a builder

using OptunaSharp;
using OptunaSharp.Visualization;

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

List<OptimizationHistoryInfo> history = OptimizationHistory.GetInfoList(study);
foreach (OptimizationHistoryInfo info in history)
    // info.ValuesInfo carries the raw and best-so-far series, error bars, feasibility, etc.
    Render(info);

The builders

Each builder is a static class; its entry method returns a public sealed record info type.

Builder	Entry method	Returns	Shows
`OptimizationHistory`	`GetInfoList(study, …)`	`OptimizationHistoryInfo` (list)	Objective value and best-so-far over trials
`IntermediateValues`	`GetPlotInfo(study)`	`IntermediatePlotInfo`	Reported intermediate values per trial
`Edf`	`GetInfo(study, …)`	`EdfInfo`	Empirical distribution of objective values
`SlicePlot`	`GetPlotInfo(study, …)`	`SlicePlotInfo`	Objective vs. each parameter
`Contour`	`GetInfo(study, …)`	`ContourInfo`	Objective over parameter pairs
`Rank`	`GetInfo(study, …)`	`RankPlotInfo`	Trials ranked over parameter pairs
`ParallelCoordinate`	`GetInfo(study, …)`	`ParallelCoordinateInfo`	All params + objective as parallel axes
`ParamImportances`	`GetInfo` / `GetInfos(study, …)`	`ImportancesInfo`	Parameter importances as bars
`ParetoFront`	`GetInfo(study, …)`	`ParetoFrontInfo`	The Pareto front (multi-objective)
`HypervolumeHistory`	`GetInfo(study, referencePoint)`	`HypervolumeHistoryInfo`	Dominated hypervolume over trials
`Timeline`	`GetInfo(study, nRecentTrials?)`	`TimelineInfo`	Trial start/finish timeline
`TerminatorImprovement`	`GetInfo(study, …)`	`ImprovementInfo`	Improvement vs. error over trials

Most builders accept optional @params (which parameters to include), target (Func<FrozenTrial, double> — pick an objective in a multi-objective study), and targetName (the axis label). HypervolumeHistory requires a referencePoint; SlicePlot and IntermediateValues use GetPlotInfo rather than GetInfo.

Building your own chart

The info records expose everything a renderer needs. For example, OptimizationHistoryInfo carries the trial numbers, the raw ValuesInfo series, the best-so-far series, optional error bars, and a per-point feasibility/completeness ValueState — enough to draw the standard history chart in any library, or to export as CSV/JSON.

Relationship to Optuna

Optuna (Python)	OptunaSharp
`optuna.visualization._optimization_history._get_optimization_history_info_list(study)`	`OptimizationHistory.GetInfoList(study)`
`_get_pareto_front_info(study)`	`ParetoFront.GetInfo(study)`
`_get_hypervolume_history_info(study, ref_point)`	`HypervolumeHistory.GetInfo(study, referencePoint)`
`plot_optimization_history(study)` (rendering)	`PlotlyVisualization.PlotOptimizationHistory(study)` in the Plotly satellite

The info builders are bit-exact against live Optuna wherever the path avoids the error function (and within the documented erf tolerance where it doesn't).

Divergences. ParallelCoordinate may compute the line draw order differently on tied ranks (cosmetic — it only affects which lines paint first); Rank's RGB color values come from Plotly's colorscale sampling and are deferred to the rendering layer (the data layer exposes the pure color indices and an infeasibility mask instead).

Table of Contents

Visualization

Using a builder

The builders

Building your own chart

Relationship to Optuna

See also