Crate sheaf 

Hierarchical abstraction: tree structures + reconciliation / conformal primitives for multi-resolution views.

Default build is hierarchy-first (minimal dependencies). Algorithmic clustering and community detection are opt-in via feature flags.

Re-exports

pub use crate::cluster::ItDendrogram;

pub use crate::hierarchy::HierarchicalConformal;

pub use crate::hierarchy::HierarchyTree;

pub use crate::learnable_sheaf::LearnableSheaf;

pub use crate::learnable_sheaf::RestrictionFamily;

pub use crate::reconciliation::reconcile;

pub use crate::reconciliation::ReconciliationMethod;

pub use crate::reconciliation::SummingMatrix;

pub use crate::sheaf_laplacian::CellularSheaf;

pub use error::Error;

pub use error::Result;

pub use metrics::ari;

pub use metrics::completeness;

pub use metrics::fowlkes_mallows;

pub use metrics::homogeneity;

pub use metrics::nmi;

pub use metrics::purity;

pub use metrics::v_measure;

pub use distribution_distance::DistributionDistance;

pub use distribution_distance::DistributionDistanceConfig;

pub use cluster::Clustering;

pub use cluster::Gmm;

pub use cluster::HierarchicalClustering;

pub use cluster::Kmeans;

pub use cluster::Linkage;

pub use cluster::SoftClustering;

pub use community::CommunityDetection;

pub use community::LabelPropagation;

pub use community::Leiden;

pub use community::Louvain;

pub use community::knn_graph_from_embeddings;

pub use community::knn_graph_with_config;

pub use community::KnnGraphConfig;

pub use community::WeightFunction;

pub use summarize::Summarizer;

pub use hierarchy::Dendrogram;

pub use hierarchy::RaptorTree;

pub use hierarchy::TreeConfig;

Modules

cluster

Clustering algorithms for grouping similar items.

community

Community detection algorithms for graphs.

distribution_distance

Distribution-distance utilities for comparing clusters as point clouds.

error

Error types used across sheaf.

hierarchy

Hierarchical structures for multi-resolution retrieval.

learnable_sheaf

Learnable (parametric) restriction maps for the sheaf Laplacian.

metrics

Clustering evaluation metrics.

reconciliation

Reconciliation methods for hierarchical forecasts.

sheaf_laplacian

Cellular sheaf Laplacian on graphs.

summarize

Summarization strategies for hierarchical abstraction.

Structs

CopKmeans

COP-Kmeans: constrained k-means clustering (Wagstaff et al., 2001).

CorrelationClustering

Correlation clustering via PIVOT with optional local search refinement.

CorrelationResult

Result of correlation clustering.

CosineDistance

Cosine distance: 1 - cos_sim(a, b).

DenStream

DenStream: streaming density-based clustering.

Euclidean

Euclidean (L2) distance: sqrt(sum((a_i - b_i)^2)).

FlatRef

Zero-copy view into a contiguous row-major f32 buffer.

MiniBatchKmeans

Mini-Batch K-means clustering (Sculley, 2010).

SignedEdge

A signed edge between two items with a similarity score.

SquaredEuclidean

Squared Euclidean distance: sum((a_i - b_i)^2).

Enums

Constraint

A pairwise constraint for semi-supervised clustering.

Traits

DataRef

Trait for read-only access to a 2D dataset of f32 rows.

DistanceMetric

A distance function between two equal-length f32 slices.