Skip to main content

umap_from_knn()

Run UMAP on a precomputed k-nearest-neighbor graph.

Usage

Source

umap_from_knn(
    indices,
    distances,
    *,
    n_components=2,
    min_dist=0.1,
    spread=1.0,
    n_epochs=None,
    learning_rate=1.0,
    init="auto",
    negative_sample_rate=5,
    repulsion_strength=1.0,
    local_connectivity=1.0,
    set_op_mix_ratio=1.0,
    a=None,
    b=None,
    random_state=None,
    optimizer=None,
    layout="0.6",
    disconnection_distance=None,
    n_threads=None,
    gpu=False,
    verbose=False
)

Use this when you already have neighbors, for example from knn(), an exact search, a vector database or FAISS. The original vectors are not needed.

Parameters

indices: IntoExpr

Array or List column of integer row indices of each row’s nearest neighbors, sorted by ascending distance. The row itself may be included (usually first, at distance 0) or left out. Every row must have the same number of neighbors.

distances: IntoExpr

Array or List column with the matching distances.

n_components: int = 2
min_dist: int = 2
spread: int = 2
n_epochs: int = 2
learning_rate: int = 2
negative_sample_rate: int = 2
repulsion_strength: float = 1.0
local_connectivity: float = 1.0
set_op_mix_ratio: float = 1.0
a: float = 1.0
b: float = 1.0
random_state: float = 1.0
optimizer: Optimizer | None = None

See umap().

layout: Optimizer | None = None

See umap().

disconnection_distance: Optimizer | None = None

See umap().

n_threads: Optimizer | None = None

See umap().

gpu: Optimizer | None = None

See umap().

verbose: Optimizer | None = None

See umap().

init: InitLike = "auto"
See umap(). "pca" is not available because it needs the vectors.

Returns

pl.Expr
An Array(Float32, n_components) expression.