umap-learn 0.5.4+dfsg-1 (i386 binary) in ubuntu oracular

 Uniform Manifold Approximation and Projection (UMAP) is a dimension
 reduction technique that can be used for visualisation similarly to t-
 SNE, but also for general non-linear dimension reduction. The algorithm
 is founded on three assumptions about the data:
 .
  1. The data is uniformly distributed on a Riemannian manifold;
  2. The Riemannian metric is locally constant (or can be
     approximated as such);
  3. The manifold is locally connected.
 .
 From these assumptions it is possible to model the manifold with a fuzzy
 topological structure. The embedding is found by searching for a low
 dimensional projection of the data that has the closest possible
 equivalent fuzzy topological structure.