nb_filter: Boosting via C++ function. Parallelisation by R-package...

In netboost: Network Analysis Supported by Boosting

Description

Parallelisation via multicore (via 'parallel'-package). So *nix only atm.

Usage

nb_filter(datan, stepno = 20L, until = 0L, progress = 1000L,
  filter_method = c("boosting", "skip", "kendall", "spearman",
  "pearson"), cores = getOption("mc.cores", 2L), mode = 2L,
  verbose = getOption("verbose"))

Arguments

datan	Data frame were rows correspond to samples and columns to features.
stepno	Integer amount of boosting steps
until	Stop at index/column (if 0: iterate through all columns)
progress	Integer. If > 0, print progress after every X steps (mind: parallel!)
filter_method	The following filtering methods are supported: "boosting" (non-zero coefficients in likelihood based boosting), "skip" (no filter), "kendall" (stats::cor.test), "spearman" (stats::cor.test), "pearson" (stats::cor.test)
cores	Integer. Amount of CPU cores used (<=1 : sequential)
mode	Integer. Mode (0: x86, 1: FMA, 2: AVX). Features are only available if compiled accordingly and available on the hardware.
verbose	Additional diagnostic messages.

Value

matrix n times 2 matrix with the indicies of the n unique entrees of the filter

Examples

data('tcga_aml_meth_rna_chr18',  package='netboost')
 cores <- as.integer(getOption('mc.cores', 2))
 datan <- as.data.frame(scale(tcga_aml_meth_rna_chr18, center=TRUE,
 scale=TRUE))
 filter <- nb_filter(datan=datan, stepno=20L, until=0L, progress=1000L,
 cores=cores,mode=2L)
 head(filter)
 nrow(filter)/(ncol(datan)*(ncol(datan)-1)/2) # proportion of potential undirected edges

netboost documentation built on Nov. 8, 2020, 4:58 p.m.