rust-ml/linfa

Provides a modular ecosystem of specialized algorithm crates spanning supervised learning (linear/logistic regression, SVMs, decision trees), unsupervised methods (K-means, DBSCAN, PCA, t-SNE), and preprocessing utilities, with optional BLAS/LAPACK acceleration via OpenBLAS, Netlib, or Intel MKL. Built as composable sub-packages inspired by scikit-learn's API design, enabling users to mix classical algorithms for domain-specific workflows without monolithic dependencies.

4,579 stars. Actively maintained with 3 commits in the last 30 days.