Gaussian Mixture Model
Probabilistic soft clustering with per-component Gaussian distributions
Gaussian Mixture Model (GMM) models the data as a mixture of Gaussian distributions. Unlike K-Means, it produces soft assignments — the probability that each point belongs to each component.
When to use:
- When cluster membership probabilities (uncertainty) are needed
- Ellipsoidal clusters of varying size and orientation
- Density estimation alongside clustering
Input: Tabular data with the feature columns defined during training Output: Most likely cluster label and component probabilities per row
Model Settings (set during training, used at inference)
N Components (default: 1) Number of mixture components (clusters). Tune during training using BIC or AIC.
Covariance Type (default: full) Shape of cluster covariance:
full— each component has its own full covariance matrix (most flexible)tied— all components share one covariance matrixdiag— diagonal covariance per componentspherical— single variance per component
Tol (default: 0.001) EM convergence threshold.
Max Iter (default: 100) Maximum EM iterations.
Inference Settings
No dedicated inference-time settings. The trained Gaussian parameters determine soft cluster assignments.