cyanea-ml

Machine learning primitives for bioinformatics: clustering, classification, dimensionality reduction, and sequence encoding.

269 unit tests + 4 doc tests.

What's Inside

• Clustering -- K-means, DBSCAN, hierarchical (single/complete/average/Ward linkage)
• Distance metrics -- Euclidean, Manhattan, cosine, Hamming; pairwise distance matrices
• KNN -- k-nearest neighbors classification and regression
• Linear regression -- ordinary least squares via normal equation
• Decision trees -- CART classifier with Gini impurity
• Random forests -- bagged ensemble with bootstrap sampling and feature bagging
• Gradient boosted trees -- regression, binary/multiclass classification, early stopping, feature importance
• Feature selection -- variance threshold, mutual information, recursive feature elimination, Lasso L1
• Hidden Markov models -- forward, backward, Viterbi, Baum-Welch EM
• PCA -- principal component analysis (with optional BLAS acceleration)
• t-SNE -- t-distributed stochastic neighbor embedding
• UMAP -- uniform manifold approximation and projection
• Sequence encoding -- one-hot and label encoding for DNA/RNA/protein
• K-mer counting -- frequency vectors for ML features
• Sequence embeddings -- k-mer frequency and composition vector embeddings
• Classification metrics -- confusion matrix, precision/recall/F1, ROC/PR curves, AUC, MCC
• Cross-validation -- k-fold, stratified k-fold, leave-one-out
• Normalization -- min-max, z-score, L2 (sample-wise and column-wise)
• Evaluation -- silhouette score for cluster quality

Quick Start

[dependencies]
cyanea-ml = "0.1"

use cyanea_ml::{
    cluster::{kmeans, KMeansConfig},
    distance::pairwise_distances,
    distance::DistanceMetric,
};

let data = vec![1.0, 2.0, 10.0, 11.0, 20.0, 21.0]; // 3 points, 2 features
let config = KMeansConfig { k: 2, max_iterations: 100, tolerance: 1e-4 };
let result = kmeans(&data, config).unwrap();
println!("Labels: {:?}", result.labels);

Feature Flags

Modules

See Also

• API Reference
• Usage Guide
• Architecture