RunjrSiCKLSNMF: Run jrSiCKLSNMF on an object of class SickleJr

Description

Wrapper function to run jrSiCKLSNMF on an object of class SickleJr. Performs jrSiCKLSNMF on the given SickleJr

Usage

RunjrSiCKLSNMF(
  SickleJr,
  rounds = 30000,
  differr = 1e-06,
  display_progress = TRUE,
  lossonsubset = FALSE,
  losssubsetsize = dim(SickleJr@H)[1],
  minibatch = FALSE,
  batchsize = 1000,
  random_W_updates = FALSE,
  seed = NULL,
  minrounds = 200,
  suppress_warnings = FALSE,
  subsample = 1:dim(SickleJr@normalized.count.matrices[[1]])[2]
)

Value

An object of class SickleJr with updated \(\mathbf{W}^v\) matrices, updated \(\mathbf{H}\) matrix, and a vector of values for the loss function added to the Wlist, H, and loss slots, respectively

Arguments

SickleJr: An object of class SickleJr
rounds: Number of rounds: defaults to 2000
differr: Tolerance for percentage change in loss between updates: defaults to 1e-6
display_progress: Boolean indicating whether to display the progress bar for jrSiCKLSNMF
lossonsubset: Boolean indicating whether to use a subset to calculate the loss function rather than the whole dataset
losssubsetsize: Size of the subset of data on which to calculate the loss
minibatch: Boolean indicating whether to use mini-batch updates
batchsize: Size of batch for mini-batch updates
random_W_updates: Boolean indicating whether or not to use random_W_updates updates (i.e. only update \(\mathbf{W}^v\) once per mini-batch epoch)
seed: Number specifying desired random seed
minrounds: Minimum number of rounds: most helpful for the mini-batch algorithm
suppress_warnings: Boolean indicating whether to suppress warnings
subsample: A numeric used primarily when finding an appropriate number of latent factors: defaults to total number of cells

References

Cai2008jrSiCKLSNMF

jnmf2009jrSiCKLSNMF

Eddelbuettel2011jrSiCKLSNMF

Eddelbuettel2014jrSiCKLSNMF

Elyanow2020jrSiCKLSNMF

halfbakednmfjrSiCKLSNMF

Lee1999jrSiCKLSNMF

Liu2013jrSiCKLSNMF

Examples

Run this code

SimSickleJrSmall<-RunjrSiCKLSNMF(SimSickleJrSmall,rounds=5)

Run the code above in your browser using DataLab