trainSlalom: Train a SlalomModel object

In PMBio/Rslalom: Factorial Latent Variable Modeling of Single-Cell RNA-Seq Data

Description

Train a SlalomModel to infer model parameters.

Usage

trainSlalom(object, nIterations = 5000, minIterations = 700,
  tolerance = 1e-08, forceIterations = FALSE, shuffle = TRUE,
  pretrain = TRUE, verbose = TRUE, seed = NULL, drop_factors = TRUE)

Arguments

object	a Rcpp_SlalomModel object
nIterations	integer(1) maximum number of iterations to use in training the model (default: 5000)
minIterations	integer(1) minimum number of iterations to perform.
tolerance	numeric(1) tolerance to allow between iterations (default 1e-08)
forceIterations	logical(1) should the model be forced to update nIteration times?
shuffle	logical(1) should the order in which factors are updated be shuffled between iterations? Shuffling generally helps speed up convergence so is recommended and defaults is TRUE
pretrain	logical(1), should the model be "pre-trained" to achieve faster convergence and obtain an initial update order? Recommended; default is TRUE
verbose	logical(1), should messages be printed about what the function is doing? Default is TRUE.
seed	integer(1) value supplying a random seed to make results reproducible (default is NULL)
drop_factors	logical(1), should factors be dropped from the model if the model determines them not to be relevant? Default is TRUE.

Details

Train the model using variational Bayes methods to infer parameters.

Value

an 'Rcpp_SlalomModel' object

Author(s)

Davis McCarthy

Examples

gmtfile <- system.file("extdata", "reactome_subset.gmt", package = "slalom")
genesets <- GSEABase::getGmt(gmtfile)
data("mesc")
model <- newSlalomModel(mesc, genesets, n_hidden = 5, min_genes = 10)
model <- initSlalom(model)
model <- trainSlalom(model, nIterations = 10)

PMBio/Rslalom documentation built on May 28, 2019, 2:23 p.m.