Sweep k on the GPU over meta cells — rs_nmf_k_sweep_mc

GPU counterpart of bixverse::rs_nmf_k_sweep_mc(). Returns diagnostics only, no factors. The matrix is uploaded once and serves every one of the length(k_range) * n_runs solves, which is where the GPU path pays off.

Usage

rs_nmf_k_sweep_mc_gpu(
  sparse_data,
  k_range,
  preprocessing,
  use_second_layer,
  nmf_hals_params,
  nmf_consensus_params,
  n_runs,
  seed,
  verbose
)

Arguments

sparse_data: A named list with data, indptr, indices, nrow, ncol and cs_type.
k_range: Integer vector. Ranks to evaluate, every entry at least 2 and at most 128, the GPU solver's rank cap.
preprocessing: String. One of c("none", "sd", "sqrt_sd").
use_second_layer: Boolean. If TRUE, runs NMF on normalised counts.
nmf_hals_params: Named list. Contains the NMF parameters.
nmf_consensus_params: Named list. Contains the consensus parameters.
n_runs: Integer. Number of restarts per k. Must be at least 2.
seed: Integer. Base random seed.
verbose: Integer. 0L - quiet; 1L - normal verbosity; 2L - detailed verbosity.

Value

A list of equal-length vectors, one element per swept k: k, stability, best_error, median_error, consensus_failed, n_dropped, n_empty_clusters and n_converged.

References

Kotliar et al., eLife, 2019