GMMDetector<T>

Models & Types · AiDotNet.AnomalyDetection.Probabilistic

Detects anomalies using Gaussian Mixture Models (GMM).

Inheritance: ModelBase<T, Matrix<T>, Vector<T>> → AnomalyDetectorBase<T> → GMMDetector<T>

Implements: IGeneratedParameterRegistrar<T>

For Beginners

GMM models data as a mixture of several Gaussian distributions. Points with low probability under this model are anomalies - they don’t fit well into any of the learned Gaussian clusters.

How It Works

The algorithm works by:

  1. Fit a mixture of k Gaussians using Expectation-Maximization
  2. For each point, compute its probability under the mixture model
  3. Points with low probability are anomalies

When to use:

  • Data is a mixture of Gaussian clusters
  • Different clusters have different shapes/sizes
  • You need probabilistic anomaly scores

Industry Standard Defaults:

  • Components: 3 (typical range 2-10)
  • Covariance type: diagonal (uses variance per feature for stability)
  • Contamination: 0.1 (10%)

Constructors

Constructor Summary
GMMDetector(int nComponents, int maxIterations, double contamination, int randomSeed) Creates a new GMM anomaly detector.

Properties

Property Summary
int NComponents { get; } Gets the number of Gaussian components.

Methods

Method Summary
Vector<T> ScoreAnomalies(Matrix<T> X)
void Fit(Matrix<T> X)