IcaDecomposition<T>

Models & Types · AiDotNet.DecompositionMethods.MatrixDecomposition

Implements Independent Component Analysis (ICA) for blind source separation.

Inheritance: MatrixDecompositionBase<T> → IcaDecomposition<T>

For Beginners

ICA is a technique used to separate mixed signals into their independent sources. Imagine you’re at a party with multiple people talking, and you have multiple microphones recording the mixed conversations. ICA helps you separate out each individual voice.

How It Works

The key idea is that ICA finds statistically independent components that were combined together. This is different from other decomposition methods because it focuses on statistical independence rather than uncorrelated patterns.

Common applications include:

  • Audio source separation (cocktail party problem)
  • Brain signal analysis (EEG, fMRI)
  • Image separation and feature extraction
  • Financial data analysis
  • Telecommunications signal processing

Constructors

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

Properties

Property Summary
Matrix<T> IndependentComponents { get; set; } Gets the independent components (separated sources).
Matrix<T> MixingMatrix { get; set; } Gets the mixing matrix (inverse of unmixing matrix).
Matrix<T> UnmixingMatrix { get; set; } Gets the unmixing matrix (separation matrix).
Matrix<T> WhiteningMatrix { get; set; } Gets the whitening matrix used in the preprocessing step.
Vector<T> Mean { get; set; } Gets the mean vector used for centering the data.

Methods

Method Summary
Matrix<T> Transform(Matrix<T> X) Transforms new data using the learned ICA model.
Vector<T> Solve(Vector<T> b) Solves a linear system Ax = b using the ICA decomposition.