Build the graph + initial embedding (eager) and return a stateful [`Umap`]. The embedding and kNN graph are valid immediately at epoch 0; call [`Umap

* Wrap an embedding pipeline in the standard coalesce → batch → embed stack. * Coalesce sits on the outside so concurrent callers fold into one * ch

Build a Rust benchmark binary in release mode.

Return all cached entries as a dict of {name: value}.

Cholesky factorization of m×m SPD matrix (row-major flat). Returns lower triangle L such that A = L L^T, or None if not SPD.

* Remove orphaned entries from state. Returns the original object unchanged if no cleanup is needed. * - Charts not referenced by any layout's chartI

Create a built-in embedder function by name.

* Wraps a Readable store so that subscribers are only notified when the * value actually changes by strict reference equality (`!==`). * * Svelte's

Convert max-heap to ascending sorted order (in-place). This is the second half of heapsort.

Early stopping threshold. Default: 0.001.

* SQL expression for the list of feature names (VARCHAR[]), regardless of column * kind. For struct-list columns the names are projected out of each

Solve the generalized eigenvalue problem G*c = λ*S*c. Uses Cholesky reduction: S = L*L^T, then solve eigh(L^{-1} G L^{-T}). Returns (eigenvalues ascen

Generate MNIST benchmark datasets.

Generate a standalone HTML report from benchmark results CSV. Args: csv_path: Path to results.csv output_path: Path to write repo

Returns (fast_distance_func, correction_func) for metrics that have optimized alternatives

Return the actual point indices of the top-k highest-degree points from a subset.

Compute self^T * other → (self.ncols() × other.ncols()) matrix, row-major flat.

* Discard any in-flight run immediately. {@link extractSegments} captured live * `Text` nodes plus offsets computed against their content at that i

Convert a JS callback `(progress: number, stage: string) => void` into a Rust `ProgressCallback`.

LOBPCG for the smallest eigenvalues of the normalized Laplacian. Finds the k smallest non-trivial eigenvalues by constraining the search space to be

Build a single dense RP tree.

Build a hub tree that uses neighbor graph information for better splits.