In sysbiomed/ransac: RANSAC

knitr::opts_chunk$set(echo = TRUE)

devtools::load_all('.')
library(futile.logger)
library(parallel)
library(ggplot2)
library(reshape2)
flog.layout(layout.format('~m'))

Sampling functions in RANSAC family

# generate data
my.data <- gen.synth(obs = 100, 
                     new.coef = c(0, 3, -2, 1))

# number of sampling iterations
k <- 10000

# generate from glm sampling
my.data.sample <- mclapply(seq(k), function(ix) { 
  ransac.binomial.glm()$sample(my.data$ydata$logit_class, n = 30) 
  }, mc.cores = 10)

# table of counts
my.counts <- as.vector(table(unlist(my.data.sample)))
flog.info('Summary of counts', summary(my.counts), capture = T)

# plot
ggplot(melt(unlist(my.data.sample))) +
  geom_freqpoly(aes(value), binwidth = 1) + 
  lims(x = c(1,100)) + theme_minimal() +
  scale_y_continuous(expand = c(.5, .5), limits = c(min(my.counts), max(my.counts)))

# generate from glmnet sampling (same function as before)
my.data.sample <- mclapply(seq(k), function(ix) { 
  ransac.binomial.glm()$sample(my.data$ydata$logit_class, n = 30) 
  }, mc.cores = 10)

# table of counts
my.counts <- as.vector(table(unlist(my.data.sample)))
flog.info('Summary of counts', summary(my.counts), capture = T)

# plot
ggplot(melt(unlist(my.data.sample))) +
  geom_freqpoly(aes(value), binwidth = 1) + 
  lims(x = c(1,100)) + theme_minimal() +
  scale_y_continuous(expand = c(.5, .5), limits = c(min(my.counts), max(my.counts)))

sysbiomed/ransac documentation built on May 14, 2019, 12:55 a.m.