sklears-mixture

Latest release: 0.1.0-alpha.1 (October 13, 2025). See the workspace release notes for highlights and upgrade guidance.

Overview

sklears-mixture implements Gaussian Mixture Models, Bayesian mixtures, Dirichlet process mixtures, and clustering utilities consistent with scikit-learn’s mixture module.

Key Features

• Algorithms: GaussianMixture, BayesianGaussianMixture, DirichletProcessGaussianMixture, and spherical/covariance options.
• Inference: Expectation-Maximization, variational inference, and online updates for streaming data.
• Accelerated Kernels: SIMD and GPU-accelerated responsibilities, log-likelihood evaluation, and sampling.
• Integration: Compatible with preprocessing, model selection, and inspection crates for pipeline workflows.

Quick Start

use sklears_mixture::GaussianMixture;
use scirs2_core::ndarray::Array2;

let x: Array2<f64> = // load or generate data
    Array2::zeros((1000, 5));

let gmm = GaussianMixture::builder()
    .n_components(4)
    .covariance_type("full")
    .max_iter(200)
    .tol(1e-3)
    .random_state(Some(42))
    .build();

let fitted = gmm.fit(&x)?;
let labels = fitted.predict(&x)?;

Status

• Fully covered by the 10,013 passing workspace tests for 0.1.0-alpha.1.
• Achieves 5–15× speedups over scikit-learn on medium-sized datasets.
• Planned features (GPU variational inference, streaming DPGMM) tracked in TODO.md.