Class FoldResult<T, TInput, TOutput>

Namespace

AiDotNet.Models.Results

Assembly

AiDotNet.dll

Represents the results of a single fold in cross-validation.

public class FoldResult<T, TInput, TOutput>

Type Parameters

T

The numeric type used for calculations (e.g., float, double).

TInput

The type of input data (e.g., Matrix<T> for tabular data, Tensor<T> for images).

TOutput

The type of output data (e.g., Vector<T> for predictions, custom types for other formats).

Inheritance

object

FoldResult<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

For Beginners: A FoldResult contains all the performance metrics for one "fold" in cross-validation. Think of it like a report card for a single test of your model, where the model was trained on one subset of your data and tested on another.

Constructors

FoldResult(int, Vector<T>, Vector<T>, Vector<T>, Vector<T>, Dictionary<string, T>?, TimeSpan?, TimeSpan?, int, IFullModel<T, TInput, TOutput>?, ClusteringMetrics<T>?, int[]?, int[]?)

Creates a new instance of the FoldResult class.

public FoldResult(int foldIndex, Vector<T> trainingActual, Vector<T> trainingPredicted, Vector<T> validationActual, Vector<T> validationPredicted, Dictionary<string, T>? featureImportance = null, TimeSpan? trainingTime = null, TimeSpan? evaluationTime = null, int featureCount = 0, IFullModel<T, TInput, TOutput>? model = null, ClusteringMetrics<T>? clusteringMetrics = null, int[]? trainingIndices = null, int[]? validationIndices = null)

Parameters

foldIndex int

The index of this fold.

trainingActual Vector<T>

The actual values in the training dataset.

trainingPredicted Vector<T>

The predicted values for the training dataset.

validationActual Vector<T>

The actual values in the validation dataset.

validationPredicted Vector<T>

The predicted values for the validation dataset.

featureImportance Dictionary<string, T>

Optional dictionary of feature importance scores.

trainingTime TimeSpan?

Time taken to train the model.

evaluationTime TimeSpan?

Time taken to evaluate the model.

featureCount int

The number of features used in the model.

model IFullModel<T, TInput, TOutput>

Optional trained model instance for this fold.

clusteringMetrics ClusteringMetrics<T>

Optional clustering quality metrics for this fold.

trainingIndices int[]

Optional array of indices for training samples in this fold.

validationIndices int[]

Optional array of indices for validation samples in this fold.

Remarks

For Beginners: This constructor creates a complete report of how well your model performed on one fold of cross-validation. It calculates various error metrics and statistics that help you understand your model's strengths and weaknesses. If your model performs clustering, you can also include clustering-specific metrics.

Properties

ActualValues

Gets the actual values from the validation dataset.

public Vector<T> ActualValues { get; }

Property Value

Vector<T>

ClusteringMetrics

Gets the clustering quality metrics for this fold, if applicable.

public ClusteringMetrics<T>? ClusteringMetrics { get; }

Property Value

ClusteringMetrics<T>

Remarks

For Beginners: This property stores clustering quality metrics when your model produces cluster assignments (like K-Means, DBSCAN, or other clustering algorithms). These metrics help you understand how well the clustering performed for this specific fold:

• Silhouette Score: How well each point fits in its cluster
• Calinski-Harabasz Index: How well-separated and compact clusters are
• Davies-Bouldin Index: Average similarity between clusters (lower is better)
• Adjusted Rand Index: Similarity to ground truth labels (if available)

This property will be null if:

• The model doesn't produce cluster labels (e.g., regression or standard classification)
• Clustering metrics couldn't be calculated for this fold

When cross-validating clustering algorithms, these metrics are automatically calculated for each fold, allowing you to see how consistent your clustering is across different data splits.

EvaluationTime

Gets the time taken to evaluate the model for this fold.

public TimeSpan EvaluationTime { get; }

Property Value

TimeSpan

FeatureImportance

Gets the feature importance scores for this fold.

public Dictionary<string, T> FeatureImportance { get; }

Property Value

Dictionary<string, T>

FoldIndex

Gets the index of this fold in the cross-validation process.

public int FoldIndex { get; }

Property Value

int

Model

Gets the trained model instance for this fold.

public IFullModel<T, TInput, TOutput>? Model { get; }

Property Value

IFullModel<T, TInput, TOutput>

Remarks

For Beginners: This property stores the model that was trained specifically on this fold's training data. Having access to the individual fold models allows you to: - Analyze how different models vary across folds - Use ensemble methods by combining predictions from multiple fold models - Investigate which features are important in different data subsets

PredictedValues

Gets the predicted values for the validation dataset.

public Vector<T> PredictedValues { get; }

Property Value

Vector<T>

TrainingErrors

Gets the error statistics for the training data.

public ErrorStats<T> TrainingErrors { get; }

Property Value

ErrorStats<T>

TrainingIndices

Gets the indices of the training samples in this fold.

public int[]? TrainingIndices { get; }

Property Value

int[]

Remarks

For Beginners: These are the row indices from the original dataset that were used for training the model in this fold. For example, if this contains [0, 1, 5, 6], it means rows 0, 1, 5, and 6 from the original data were used for training.

This is especially useful for: - Nested cross-validation where you need to extract subsets based on indices - Debugging to verify correct data splits - Advanced techniques like stratified sampling verification

TrainingTime

Gets the time taken to train the model for this fold.

public TimeSpan TrainingTime { get; }

Property Value

TimeSpan

ValidationErrors

Gets the error statistics for the validation data.

public ErrorStats<T> ValidationErrors { get; }

Property Value

ErrorStats<T>

ValidationIndices

Gets the indices of the validation samples in this fold.

public int[]? ValidationIndices { get; }

Property Value

int[]

Remarks

For Beginners: These are the row indices from the original dataset that were held out for validation in this fold. For example, if this contains [2, 3, 4], it means rows 2, 3, and 4 from the original data were used for validation (testing).

This ensures you can accurately reconstruct which samples were used for validation, even when the target values contain duplicates. This prevents data leakage and ensures correct nested cross-validation.

ValidationPredictionStats

Gets the prediction statistics for the validation data.

public PredictionStats<T> ValidationPredictionStats { get; }

Property Value

PredictionStats<T>