FeatureNoise<T>

Models & Types · AiDotNet.Augmentation.Tabular

Adds Gaussian noise to numerical features in tabular data.

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

For Beginners

This augmentation adds small random variations to your data, similar to measurement noise in real-world data. This helps models become robust to small fluctuations and prevents overfitting to exact values.

How It Works

When to use:

  • Numerical features that have natural variation
  • Small datasets where regularization is needed
  • When you want to simulate measurement uncertainty

When NOT to use:

  • Categorical features (use other augmentations)
  • Features with strict constraints (e.g., binary flags)

Constructors

Constructor Summary
FeatureNoise(double noiseStdDev, double probability, int[] featureIndices) Creates a new feature noise augmentation.

Properties

Property Summary
double NoiseStdDev { get; } Gets the standard deviation of the noise.
int[] FeatureIndices { get; set; } Gets or sets the indices of features to apply noise to.

Methods

Method Summary
IDictionary<string, object> GetParameters()