Class PreprocessingInfo<T, TInput, TOutput>

Namespace

AiDotNet.Preprocessing

Assembly

AiDotNet.dll

Stores the fitted preprocessing pipeline for inference.

public class PreprocessingInfo<T, TInput, TOutput>

Type Parameters

T

The numeric type used for calculations.

TInput

The input data type (typically Matrix<T>).

TOutput

The output data type (typically Vector<T>).

Inheritance

object

PreprocessingInfo<T, TInput, TOutput>

Inherited Members

object.Equals(object)

object.Equals(object, object)

object.GetHashCode()

object.GetType()

object.MemberwiseClone()

object.ReferenceEquals(object, object)

object.ToString()

Remarks

This class encapsulates the preprocessing state needed to transform new data during inference. It stores the fitted feature pipeline and optionally a target pipeline for inverse transformation of predictions.

For Beginners: After training, your preprocessing pipeline has "learned" things like the mean and standard deviation for scaling. This class stores all that learned information so you can apply the same transformations to new data during predictions.

Constructors

PreprocessingInfo()

Creates a new instance of PreprocessingInfo<T, TInput, TOutput>.

public PreprocessingInfo()

PreprocessingInfo(PreprocessingPipeline<T, TInput, TInput>)

Creates a new instance with the specified fitted pipeline.

public PreprocessingInfo(PreprocessingPipeline<T, TInput, TInput> pipeline)

Parameters

pipeline PreprocessingPipeline<T, TInput, TInput>

The fitted feature preprocessing pipeline.

PreprocessingInfo(PreprocessingPipeline<T, TInput, TInput>, PreprocessingPipeline<T, TOutput, TOutput>?)

Creates a new instance with both feature and target pipelines.

public PreprocessingInfo(PreprocessingPipeline<T, TInput, TInput> pipeline, PreprocessingPipeline<T, TOutput, TOutput>? targetPipeline)

Parameters

pipeline PreprocessingPipeline<T, TInput, TInput>

The fitted feature preprocessing pipeline.

targetPipeline PreprocessingPipeline<T, TOutput, TOutput>

The fitted target preprocessing pipeline.

Properties

IsFitted

Gets whether the feature pipeline has been fitted to data.

public bool IsFitted { get; }

Property Value

bool

IsTargetFitted

Gets whether the target pipeline has been fitted to data.

public bool IsTargetFitted { get; }

Property Value

bool

Pipeline

Gets or sets the fitted feature preprocessing pipeline.

public PreprocessingPipeline<T, TInput, TInput>? Pipeline { get; set; }

Property Value

PreprocessingPipeline<T, TInput, TInput>

Remarks

This pipeline transforms input features (X) before they are passed to the model. It must be fitted before use during inference.

For Beginners: This is the pipeline that transforms your input data (like scaling, encoding, and imputation) before making predictions.

TargetPipeline

Gets or sets the fitted target preprocessing pipeline.

public PreprocessingPipeline<T, TOutput, TOutput>? TargetPipeline { get; set; }

Property Value

PreprocessingPipeline<T, TOutput, TOutput>

Remarks

This optional pipeline transforms targets (y) during training and can perform inverse transformation on predictions during inference.

For Beginners: If you scaled your target values during training (like log-transforming prices), this pipeline knows how to "unscale" the predictions back to the original range.

Methods

InverseTransformPredictions(TOutput)

Inverse transforms predictions using the target pipeline.

public TOutput InverseTransformPredictions(TOutput predictions)

Parameters

predictions TOutput

The predictions to inverse transform.

Returns

TOutput

The inverse transformed predictions in original scale.

Exceptions

InvalidOperationException

Thrown if the target pipeline is not configured or doesn't support inverse transform.

TransformFeatures(TInput)

Transforms input features using the fitted pipeline.

public TInput TransformFeatures(TInput input)

Parameters

input TInput

The input features to transform.

Returns

TInput

The transformed features.

Exceptions

InvalidOperationException

Thrown if the pipeline is not fitted.