Run NMF (HALS) on the GPU over a set of single cells and genes

GPU counterpart of bixverse::rs_nmf_single_sc(). The counts are loaded from the binary store once, uploaded to the device, and the whole HALS loop runs there. The NNDSVD initialisation stays on the host.

Usage

rs_nmf_single_sc_gpu(
  f_path_gene,
  gene_indices,
  cell_indices,
  k,
  preprocessing,
  use_second_layer,
  nmf_hals_params,
  seed,
  verbose
)

Arguments

f_path_gene

Path to the counts_genes.bin file.

gene_indices

Integer vector. 0-indexed(!) positions of the genes to include.

cell_indices

Integer vector. 0-indexed(!) positions of cells to include in the analysis.

k

Integer. Number of latent factors to return. At most 128, the GPU solver's rank cap.

preprocessing

String. One of c("none", "sd", "sqrt_sd"). Takes the data as is, or scales by standard deviation or squared standard deviation per feature.

use_second_layer

Boolean. If TRUE, runs NMF on the normalised counts; if FALSE, on the raw counts.

nmf_hals_params

Named list. Contains the NMF parameters.

seed

Integer. Random seed for initialisation.

verbose

Integer. 0L - quiet; 1L - normal verbosity; 2L - detailed verbosity.

Value

A list with the following items

• w - The left factor matrix (n_cells x k)

• h - The right factor matrix (k x n_genes)

• final_loss - Loss at the final iteration

• n_iter - Number of iterations the algorithm run for

• converged - Did the NMF algorithm converge