trainModel,PCOSP-method {PDATK}R Documentation

Train a PCOSP Model Based on The Data the assay trainMatrix.

Description

Uses the switchBox SWAP.Train.KTSP function to fit a number of k top scoring pair models to the data, filtering the results to the best models based on the specified paramters.

Usage

## S4 method for signature 'PCOSP'
trainModel(object, numModels = 10, minAccuracy = 0.6, ...)

Arguments

object

A PCOSP object to train.

numModels

An integer specifying the number of models to train. Defaults to 10. We recommend using 1000+ for good results.

minAccuracy

A float specifying the balanced accurary required to consider a model 'top scoring'. Defaults to 0.6. Must be in the range [0, 1].

...

Fall through arguments to BiocParallel::bplapply. Use this to configure parallelization options. By default the settings inferred in BiocParallel::bpparam() will be used.

Details

This function is parallelized with BiocParallel, thus if you wish to change the back-end for parallelization, number of threads, or any other parallelization configuration please pass BPPARAM to bplapply.

Value

A PCOSP object with the trained model in the model slot.

See Also

switchBox::SWAP.KTSP.Train BiocParallel::bplapply

Examples

data(samplePCOSPmodel)

# Set parallelization settings
BiocParallel::register(BiocParallel::SerialParam())

set.seed(getModelSeed(samplePCOSPmodel))
trainModel(samplePCOSPmodel, numModels=2, minAccuracy=0.6)


[Package PDATK version 1.0.2 Index]