docs/functions/run.cv.spear.md

In jgygi/SPEAR: SPEAR - Sparse Supervised Bayesian Factor Model for Multi-omic Analyses

run.cv.spear

Run CV SPEAR on a given X and Y (stored under SPEARobject$data$train)

Description

Run CV SPEAR on a given X and Y (stored under SPEARobject$data$train)

Usage

run.cv.spear(
  fold.ids = NULL,
  num.folds = NULL,
  only.cross.Y = FALSE,
  num.cores = NULL,
  fold.id.method = "balanced",
  parallel.method = NULL,
  do.cv.eval = TRUE,
  nlambda = 100,
  calculate.factor.contributions = TRUE,
  max_iter = 10000,
  multinomial_loss = "deviance"
)

Arguments

Argument |Description ------------- |---------------- fold.ids | Assignment of folds for each subject. Must be a vector of length(N), where N = num.samples. Fold.ids must span from 1 - num.folds. Defaults to NULL (randomly assigned) num.folds | Number of folds. How many folds to use? Defaults to 5. num.cores | How many cores to use for parallel processing? Defaults to parallel::detectCores() fold.id.method | How to generate fold.ids? See ?generate.fold.ids for more information. Defaults to "balanced". parallel.method | Which parallel method to use? Can be "parLapply" (defaults when NULL), "mclapply", or "lapply" (for single threaded processing). do.cv.eval | Whether or not to automatically run $cv.evaluate(...) after constructing factors? Defaults to TRUE nlambda | For cv.evaluate(...). Number of lambdas (defaults to 100) calculate.factor.contributions | For cv.evaluate(...). Calculate factor contributions? When $params$family == "multinomial" or "ordinal" can save time to put FALSE. Defaults to TRUE. max_iter | For cv.evaluate(...). Maximum number of iterations (defaults to 10000) multinomial_loss | For cv.evaluate(...). Type of loss for when $params$family == "multinomial". Can be "deviance" (default) or "misclassification" only.cross.y | In cross validation, only use the Y? Will use all of X for fitting even if Y is missing. Defaults to TRUE

Examples

SPEARobj <- make.SPEARobject(...)

SPEARobj$run.cv.spear()

jgygi/SPEAR documentation built on July 5, 2023, 5:35 p.m.