Expand description
Dense clustering primitives.
9 clustering algorithms for dense f32 vectors, generic over pluggable
distance metrics. SIMD-accelerated (innr), with optional GPU (Metal) and
parallel (rayon) support.
Batch: Kmeans, Dbscan, Hdbscan, EVoC, CopKmeans,
CorrelationClustering.
Streaming: MiniBatchKmeans, DenStream.
Evaluation: cluster::metrics – silhouette score, Calinski-Harabasz,
Davies-Bouldin index.
Noise points from DBSCAN/HDBSCAN are labeled with the sentinel
NOISE (usize::MAX).
Re-exports§
pub use cluster::ClusterHierarchy;pub use cluster::ClusterLayer;pub use cluster::ClusterNode;pub use cluster::CompositeDistance;Deprecated pub use cluster::Constraint;pub use cluster::CopKmeans;pub use cluster::CorrelationClustering;pub use cluster::CorrelationResult;pub use cluster::CosineDistance;pub use cluster::DataRef;pub use cluster::Dbscan;pub use cluster::DenStream;pub use cluster::DistanceMetric;pub use cluster::EVoC;pub use cluster::EVoCParams;pub use cluster::Euclidean;pub use cluster::FlatRef;pub use cluster::Hdbscan;pub use cluster::HdbscanResult;pub use cluster::InnerProductDistance;Deprecated pub use cluster::Kmeans;pub use cluster::KmeansFit;pub use cluster::MiniBatchKmeans;pub use cluster::Optics;pub use cluster::OpticsResult;pub use cluster::SignedEdge;pub use cluster::SquaredEuclidean;pub use cluster::NOISE;pub use error::Error;pub use error::Result;