LpNormScaler<T>

Models & Types · AiDotNet.Preprocessing.Scalers

Scales features (columns) by dividing each element by the column’s Lp-norm.

Inheritance: TransformerBase<T, Matrix<T>, Matrix<T>> → LpNormScaler<T>

For Beginners

This scaler standardizes each feature column to have a consistent “length”.

Think of each column as an arrow (vector) in space:

  • The Lp-norm measures the “length” of this arrow
  • This scaler divides each element by the length
  • The result is an arrow pointing in the same direction but with length = 1

Different p values provide different ways to measure length:

  • p = 1: Like measuring distance by walking along city blocks
  • p = 2: Like measuring distance “as the crow flies” (straight line)
  • Higher p values: Increasingly emphasize the largest component

For example, normalizing the column [3, 4] with p = 2 (Euclidean norm):

  • The norm is sqrt(3^2 + 4^2) = sqrt(25) = 5
  • The normalized column is [3/5, 4/5] = [0.6, 0.8]
  • This new column has length 1 using the L2 norm

This is useful for:

  • Feature normalization in machine learning models
  • Ensuring consistent scaling across feature vectors
  • Applications where feature direction matters more than magnitude

How It Works

The LpNormScaler normalizes each feature (column) by dividing every element by the column’s Lp-norm. This results in each column having a unit Lp-norm. The Lp-norm is a generalization of different vector norms based on the parameter p:

  • p = 1: Manhattan (L1) norm (sum of absolute values)
  • p = 2: Euclidean (L2) norm (square root of sum of squares)
  • p = infinity: Maximum (L-infinity) norm (maximum absolute value)

Note: This scaler operates on columns (features), which is different from the Normalizer class that operates on rows (samples). Use this when you want to normalize feature vectors; use Normalizer when you want to normalize sample vectors.

Constructors

Constructor Summary
LpNormScaler(double p, int[] columnIndices) Creates a new instance of LpNormScaler.

Properties

Property Summary
Vector<T> ColumnNorms { get; } Gets the Lp-norm of each column computed during fitting.
bool SupportsInverseTransform { get; } Gets whether this transformer supports inverse transformation.
double P { get; } Gets the p parameter that defines which Lp-norm to use.

Methods

Method Summary
string[] GetFeatureNamesOut(string[] inputFeatureNames) Gets the output feature names after transformation.