sklears-discriminant-analysis

Latest release: 0.2.0 (July 14, 2026). See the workspace release notes for highlights and upgrade guidance.

Overview

sklears-discriminant-analysis implements Linear Discriminant Analysis (LDA), Quadratic Discriminant Analysis (QDA), and related subspace methods with scikit-learn compatible APIs. The crate emphasizes numerical robustness, GPU acceleration, and seamless integration with the broader sklears ecosystem.

Key Features

• Comprehensive Algorithms: LDA, QDA, shrinkage estimators, regularized discriminant analysis, and Bayesian variants.
• Performance Optimizations: SIMD-enabled linear algebra, batched matrix factorizations, and (behind the opt-in gpu feature, not enabled by default) a real sklears_core::gpu-backed GPU path (GEMM-based class-statistics, an LDA generalized-eigenvalue solve via Cholesky reduction verified against SciPy to ~1e-15, and QDA via oxicuda-solver).
• Pipeline Support: Works with sklears pipelines and model selection utilities; predict_proba is available on the fitted estimators.

Quick Start

use sklears_discriminant_analysis::LinearDiscriminantAnalysis;
use scirs2_core::ndarray::{array, Array1};

let x = array![
    [1.0, 2.0],
    [1.5, 1.8],
    [5.0, 8.0],
    [6.0, 9.0],
];
let y = Array1::from(vec![0, 0, 1, 1]);

let lda = LinearDiscriminantAnalysis::new()
    .solver("svd")
    .shrinkage(None)
    .n_components(Some(1));

let fitted = lda.fit(&x, &y)?;
let predictions = fitted.predict(&x)?;

Status

• Covered by 322 passing tests in 0.2.0 (Partial — actively evolving).
• Numerical stability validated on high-dimensional datasets using SciRS2 linear algebra backends; NumericalStability::stable_inverse now correctly handles non-symmetric matrices via scirs2_linalg::inv (previously a stub that returned NotImplemented).
• GPU acceleration now targets the sklears_core::gpu foundation (GpuBackend/GpuArray/GpuMatrixOps), replacing the previous dead scirs2_core::gpu::* path.
• Future enhancements (incremental LDA) tracked within this crate's TODO.md.