NestedCVStrategy<T>
Models & Types · AiDotNet.Evaluation.CrossValidation
Nested Cross-Validation: uses an inner CV loop for hyperparameter tuning and outer loop for evaluation.
Implements: ICrossValidationStrategy<T>
For Beginners
Nested CV is essential when you need to both tune hyperparameters AND get an unbiased estimate of model performance:
- Outer loop: Evaluates final model performance (e.g., 5-fold)
- Inner loop: Selects best hyperparameters (e.g., 3-fold within each outer fold)
How It Works
Why is this necessary? If you tune hyperparameters on the same data you evaluate on, you get optimistically biased estimates. Nested CV keeps evaluation data completely separate from hyperparameter selection.
Structure:
Constructors
| Constructor | Summary |
|---|---|
NestedCVStrategy(int outerFolds, int innerFolds, bool shuffle, int? randomSeed) |
Initializes Nested Cross-Validation. |
Properties
| Property | Summary |
|---|---|
int NumSplits { get; } |
|
string Description { get; } |
|
string Name { get; } |
Methods
| Method | Summary |
|---|---|
IEnumerable<ValueTuple<int[], int[]>> GetInnerSplits(int[] outerTrainIndices, ReadOnlySpan<T> labels) |
Gets inner CV splits for hyperparameter tuning within a specific outer fold’s training data. |
IEnumerable<ValueTuple<int[], int[]>> Split(int dataSize, ReadOnlySpan<T> labels) |