Implementation:Scikit learn Scikit learn BaseMixture

Knowledge Sources	Scikit_learn Scikit-learn Docs
Domains	Mixture Models, Density Estimation
Last Updated	2026-02-08 15:00 GMT

Overview

Concrete tool providing the abstract base class for all mixture models in scikit-learn.

Description

BaseMixture is an abstract base class that defines the common interface and implements shared methods for all mixture model classes, including GaussianMixture and BayesianGaussianMixture. It implements the Expectation-Maximization (EM) algorithm with support for multiple initializations, warm starts, convergence monitoring, and verbose output. Subclasses must implement abstract methods for the E-step, M-step, parameter initialization, and log-likelihood computation. The class inherits from DensityMixin and BaseEstimator.

Usage

BaseMixture is not used directly. Instead, use its concrete subclasses GaussianMixture or BayesianGaussianMixture. It provides the foundation for implementing custom mixture models by subclassing and implementing the required abstract methods.

Code Reference

Source Location

• Repository: scikit-learn
• File: sklearn/mixture/_base.py

Signature

class BaseMixture(DensityMixin, BaseEstimator, metaclass=ABCMeta):
    def __init__(
        self,
        n_components,
        tol,
        reg_covar,
        max_iter,
        n_init,
        init_params,
        random_state,
        warm_start,
        verbose,
        verbose_interval,
    ):

Import

from sklearn.mixture._base import BaseMixture

I/O Contract

Inputs

Name	Type	Required	Description
n_components	int	Yes	Number of mixture components.
tol	float	Yes	Convergence threshold for EM iterations.
reg_covar	float	Yes	Regularization added to diagonal of covariance matrices.
max_iter	int	Yes	Maximum number of EM iterations.
n_init	int	Yes	Number of initializations; best result is kept.
init_params	str	Yes	Initialization method for weights, means, and covariances: "kmeans", "random", "random_from_data", or "k-means++".
random_state	int or RandomState	Yes	Random state for reproducibility.
warm_start	bool	Yes	Whether to use solution of last fit as initialization for next fit.
verbose	int	Yes	Verbosity level.
verbose_interval	int	Yes	Interval between verbose log messages.

Outputs

Name	Type	Description
converged_	bool	Whether the EM algorithm converged.
n_iter_	int	Number of EM iterations performed for the best initialization.
lower_bound_	float	Lower bound value on the log-likelihood of the best fit.

Usage Examples

Basic Usage

# BaseMixture is abstract; use GaussianMixture instead:
from sklearn.mixture import GaussianMixture
import numpy as np

X = np.array([[1, 2], [1, 4], [1, 0],
              [10, 2], [10, 4], [10, 0]])

gmm = GaussianMixture(n_components=2, random_state=0).fit(X)
print(gmm.predict(X))
print(gmm.converged_)

Related Pages

• Principle:Scikit_learn_Scikit_learn_Density_Estimation