UMAP embedding

UMAP currently uses the CPU uwot backend. GPU-aware cudaverse inputs are accepted and their source device is retained in the result metadata.

Usage

cuda_umap(
  x,
  n_components = 2L,
  n_neighbors = 15L,
  min_dist = 0.1,
  metric = "euclidean",
  n_epochs = NULL,
  seed = NULL,
  ...,
  reduced_dim = NULL
)

Arguments

x

Numeric observation-by-feature matrix, compatible cudaverse result, or a SingleCellExperiment with a reduced dimension.

n_components

Output dimensions.

n_neighbors

Number of nearest neighbours.

min_dist

Minimum UMAP distance.

metric

Distance metric passed to uwot::umap().

n_epochs

Optional training epochs.

seed

Optional random seed.

...

Additional arguments passed to uwot::umap().

reduced_dim

For a SingleCellExperiment, the reduced-dimension name to embed. When NULL, a compatible recorded metadata choice is used first, followed by a uniquely named "PCA". Other names must be selected explicitly.

Value

A cuda_embedding list containing coordinates, method, backend, compute_device, per-stage compute_stages, source metadata, and algorithm parameters.

Examples

if (requireNamespace("uwot", quietly = TRUE)) {
  cuda_umap(matrix(rnorm(120), 40, 3), n_neighbors = 5, seed = 1)
}
#> <cuda_embedding method=umap observations=40 dimensions=2 backend=uwot compute_device=cpu>