seqme.utils.umap

Contents

seqme.utils.umap#

seqme.utils.umap(embeddings, seed=0)[source]#

Project embeddings into 2D using UMAP.

Parameters:

  • embeddings (ndarray | list[ndarray]) – 2D array where each row is a data point.

  • seed (int | None) – Seed for deterministic computation of UMAP.

Return type:

ndarray | list[ndarray]

Returns:

2D array of shape (n_samples, 2) or list.

Notes

UMAP is a nonlinear manifold learning method that preserves both local and some global structure, offering speed and scalability comparable to or better than t-SNE.

Contents