Home › Abstractions › The shared sparse primitive

CsrMatrix.RowL1Norm

The sum of one row's absolute values — its total mass.

public double RowL1Norm(int row)

Parameters — row is the zero-based row index.

Returns — double, the sum of |value| over that row's stored cells. Zero for an empty row.

Exceptions — IndexOutOfRangeException when row is negative or not below RowCount. The index reaches the backing array directly, so the array's own exception is what surfaces rather than a re-wrapped one.

Example — on raw counts, the L1 norm is how many terms the document holds.

using Lodestar.Abstractions;
using Lodestar.Text.Vectorization;

string[] docs = ["the cat eats", "the dog eats", "the cat and the dog"];
CsrMatrix counts = new CountVectorizer().FitTransform(docs);

double first = counts.RowL1Norm(0);   // => 3
double third = counts.RowL1Norm(2);   // => 5

Remarks — absolute values, which matters only where a matrix can hold negatives: HashingVectorizer with AlternateSign on produces them by design, and there the L1 norm is the total weight rather than a net one.

Dividing a row by this norm turns it into a distribution over terms, which is what NormalizeRows(SparseNorm.L1) does.

Applies to — net10.0, netstandard2.0.

See also — CsrMatrix.RowL2Norm, CsrMatrix.NormalizeRows, SparseNorm.