NmfDecomposition<T>

Models & Types ยท AiDotNet.DecompositionMethods.MatrixDecomposition

Implements Non-negative Matrix Factorization (NMF) for matrices with non-negative elements.

Inheritance: MatrixDecompositionBase<T> โ†’ NmfDecomposition<T>

For Beginners

NMF is a way to break down a matrix containing only non-negative values (zero or positive numbers) into two simpler matrices W and H, where V ~= W * H. Think of it like finding hidden patterns or features in your data.

How It Works

For example, if you have a matrix of movie ratings (all non-negative), NMF can discover:

  • W: How much each user likes different movie genres (features)
  • H: How much each movie belongs to different genres (features)

Common applications include:

  • Topic modeling in text documents
  • Image processing and feature extraction
  • Collaborative filtering in recommendation systems
  • Audio source separation
  • Bioinformatics data analysis

Constructors

Constructor Summary
NmfDecomposition(Matrix<T> matrix, int? components, int maxIterations, double tolerance) Initializes a new instance of the NMF decomposition for the specified matrix.

Properties

Property Summary
Matrix<T> H { get; set; } Gets the coefficient matrix H (weights/encodings).
Matrix<T> W { get; set; } Gets the basis matrix W (features/components).
int Components { get; set; } Gets the number of components (features) used in the factorization.

Methods

Method Summary
Matrix<T> Reconstruct() Reconstructs the original matrix from the factorization.
Vector<T> Solve(Vector<T> b) Solves a linear system Ax = b using the NMF decomposition.