Implementation:Dotnet Machinelearning AutoMLExperiment Configuration

Knowledge Sources	ML.NET ML.NET AutoML API
Domains	Machine_Learning, AutoML
Last Updated	2026-02-09 00:00 GMT

Overview

Concrete tool for configuring an AutoML experiment using a fluent builder API provided by ML.NET AutoML.

Description

The AutoMLExperiment class and its extension methods provide a fluent builder API for configuring all aspects of an AutoML search. The AutoCatalog.CreateExperiment() factory method instantiates the experiment. Each subsequent configuration method (SetTrainingTimeInSeconds, SetBinaryClassificationMetric, SetDataset, SetPipeline, SetTuner) returns the same AutoMLExperiment instance, enabling method chaining. The experiment accumulates configuration state and validates it when execution begins.

Usage

Import Microsoft.ML.AutoML after constructing a sweepable pipeline and loading training/validation datasets. Use the fluent API to chain all configuration calls in a single expression. The experiment must have a time budget, a metric, a dataset, and a pipeline set before execution.

Code Reference

Source Location

Repository: ML.NET
File: src/Microsoft.ML.AutoML/API/AutoCatalog.cs (Lines 308-313 for CreateExperiment)
File: src/Microsoft.ML.AutoML/AutoMLExperiment/AutoMLExperiment.cs (Lines 77-89 for SetTrainingTimeInSeconds)
File: src/Microsoft.ML.AutoML/API/AutoMLExperimentExtension.cs (Lines 86-90 for SetBinaryClassificationMetric, Lines 141-151 for SetPipeline, Lines 30-45 for SetDataset)

Signature

// Create a new AutoML experiment
public AutoMLExperiment CreateExperiment(AutoMLExperiment.AutoMLExperimentSettings settings = null)

// Set the maximum training time in seconds
public AutoMLExperiment SetTrainingTimeInSeconds(uint trainingTimeInSeconds)

// Set the optimization metric for binary classification
public static AutoMLExperiment SetBinaryClassificationMetric(
    this AutoMLExperiment experiment,
    BinaryClassificationMetric metric,
    string labelColumn = "Label",
    string predictedColumn = "PredictedLabel")

// Set the training and validation datasets
public static AutoMLExperiment SetDataset(
    this AutoMLExperiment experiment,
    IDataView train,
    IDataView validation,
    bool subSamplingTrainDataset = false)

// Set the sweepable pipeline to search
public static AutoMLExperiment SetPipeline(
    this AutoMLExperiment experiment,
    SweepablePipeline pipeline)

// Set the tuner algorithm for hyperparameter search
public AutoMLExperiment SetTuner<TTuner>()
    where TTuner : class, ITuner

Import

using Microsoft.ML.AutoML;

I/O Contract

Inputs

Name	Type	Required	Description
trainingTimeInSeconds	`uint`	Yes	Maximum wall-clock time for the experiment in seconds
metric	`BinaryClassificationMetric`	Yes	Optimization metric (Accuracy, AUC, AreaUnderPrecisionRecallCurve, F1Score, PositivePrecision, PositiveRecall, NegativePrecision, NegativeRecall)
labelColumn	`string`	No (default: "Label")	Name of the label column
predictedColumn	`string`	No (default: "PredictedLabel")	Name of the predicted label column
train	`IDataView`	Yes	Training dataset
validation	`IDataView`	Yes	Validation dataset for metric evaluation
subSamplingTrainDataset	`bool`	No (default: false)	Whether to sub-sample training data for faster early trials
pipeline	`SweepablePipeline`	Yes	The pipeline search space to explore
settings	`AutoMLExperimentSettings`	No	Optional initial settings object

Outputs

Name	Type	Description
(return)	`AutoMLExperiment`	The same experiment instance (fluent builder), fully configured and ready for execution

Usage Examples

Basic Example

var mlContext = new MLContext();

// Split data into train/validation sets
var trainTestSplit = mlContext.Data.TrainTestSplit(data, testFraction: 0.2);

// Build the sweepable pipeline
var pipeline = mlContext.Auto()
    .Featurizer(trainTestSplit.TrainSet, outputColumnName: "Features")
    .Append(mlContext.Auto().BinaryClassification(
        labelColumnName: "Label",
        featureColumnName: "Features"));

// Configure the AutoML experiment using the fluent API
var experiment = mlContext.Auto()
    .CreateExperiment()
    .SetTrainingTimeInSeconds(120)
    .SetBinaryClassificationMetric(BinaryClassificationMetric.AUC, labelColumn: "Label")
    .SetDataset(trainTestSplit.TrainSet, trainTestSplit.TestSet)
    .SetPipeline(pipeline);

With Custom Tuner Example

// Configure with a specific tuner algorithm
var experiment = mlContext.Auto()
    .CreateExperiment()
    .SetTrainingTimeInSeconds(300)
    .SetBinaryClassificationMetric(BinaryClassificationMetric.F1Score)
    .SetDataset(trainData, validationData)
    .SetPipeline(pipeline)
    .SetTuner<GridSearchTuner>();

Related Pages

Implements Principle

Principle:Dotnet_Machinelearning_Experiment_Configuration

Requires Environment

Environment:Dotnet_Machinelearning_Dotnet_SDK_And_Runtime

Uses Heuristic

Page Connections

Double-click a node to navigate. Hold to expand connections.

Principle

Implementation

Heuristic

Environment