llv.platypus: Label learning validation

In graim/PLATYPUS: Multiple view and task learning

Description

Similar to cross-fold validation, label learning validation for platypus is used to help identify the number of iterations to run when training a platypus model, so that label learning is most effective.

Usage

llv.platypus(view.list, fn.labs, llv.folds = 10, n.iters = 100,
  majority.threshold.percent = 100, nfolds = 10,
  expanded.output = TRUE, updating = FALSE,
  ignore.label = "intermediate", parallel = FALSE, classcol.labs = 1,
  output.folder = NA)

Arguments

view.list	List of view objects
fn.labs	File containing outcome labels
llv.folds	number of folds for label learning validation (similar to cross validation folds), default=10
n.iters	Maximal number of iterations for each platypus run, default=100
majority.threshold.percent	Percent agreement required to learn a sample's class label, default=100
nfolds	Number of cross-validation folds
expanded.output	Expanded output: returned result list contains a list of trained views after each iteration, default=FALSE
updating	Updating the accuracies of the single views in each iteration, default=FALSE
ignore.label	Label class to ignore, if any. Defaults to 'intermediate'
parallel	Whether or not to run in parallel mode TODO remove? numcores enables anyway
classcol.labs	Column containing the task label data. Default 1.
output.folder	Name of the folder where output is stored.

Value

A list containing fold.accuracy, labelling.matrix,labelling.matrices.views

graim/PLATYPUS documentation built on Oct. 4, 2019, 2:05 p.m.