cut_tree_to_labels

cut_tree_to_labels(
    tree,
    n_points,
    n_clusters,
    *,
    max_depth=None,
    embeddings=None,
)

Cut a DYF tree into flat cluster labels.

Parameters

Name Type Description Default
tree Tree dict from build_dyf_tree. required
n_points Total number of points. required
n_clusters Desired number of clusters. required
max_depth Unused. Accepted so callers written against the two-tree dispatcher keep working; it was only ever required for PCA trees. None
embeddings (n, d) array — required, used for leaf centroids. None

Returns

Name Type Description
np.ndarray of shape (n_points,) with cluster labels.

Raises

Name Type Description
ValueError If the tree shape is unrecognized or embeddings is missing.