AdasynAugmenter<T>

Models & Types · AiDotNet.Augmentation.Tabular

Implements ADASYN (Adaptive Synthetic Sampling) for imbalanced datasets.

Inheritance: AugmentationBase<T, Matrix<T>> → TabularAugmenterBase<T> → AdasynAugmenter<T>

For Beginners

ADASYN is an extension of SMOTE that adaptively generates more synthetic samples in regions where the minority class is harder to learn (i.e., where there are more majority class neighbors). This focuses the synthetic data where it’s most needed.

How It Works

How ADASYN differs from SMOTE:

  • SMOTE: Generates the same number of synthetic samples for each minority instance
  • ADASYN: Generates more synthetic samples for minority instances that have more majority neighbors

This adaptive approach helps the classifier focus on the hardest-to-learn examples.

Algorithm:

  • For each minority sample, calculate the ratio of majority neighbors to total neighbors
  • Normalize these ratios to get a distribution
  • Generate synthetic samples proportionally - more for samples with more majority neighbors

When to use:

  • When the decision boundary is complex and irregular
  • When minority samples near the boundary need more representation
  • When standard SMOTE doesn’t improve minority class recall enough

Reference: He et al., “ADASYN: Adaptive Synthetic Sampling Approach for Imbalanced Learning” (2008)

Constructors

Constructor Summary
AdasynAugmenter(int kNeighbors, double beta, double probability) Creates a new ADASYN augmenter.

Properties

Property Summary
double Beta { get; } Gets the target balance ratio between minority and majority classes.
int KNeighbors { get; } Gets the number of nearest neighbors to consider.

Methods

Method Summary
IDictionary<string, object> GetParameters()
Matrix<T> GenerateSyntheticSamples(Matrix<T> minorityData, Matrix<T> majorityData, AugmentationContext<T> context) Applies ADASYN to generate synthetic samples for the minority class.