tests/testthat/test_svmmajcrossval.R

In SVMMaj: Implementation of the SVM-Maj Algorithm

library(magrittr)
library(testthat)
library(SVMMaj)
library(kernlab)

n = 200
k = 4
beta    <- matrix(rnorm(k + 1), ncol = 1)
X.base  <- matrix(rnorm(n * k), ncol = k) 
y.base  <- scale(X.base %*% beta[-1] + beta[1]) > 0

test_that(
  desc = "Only lambda changes in crossvalidation", {
    model <- svmmajcrossval(
      X.base, y.base, verbose = TRUE, 
      search.grid = list(lambda = c(0.01,0.1,1)),
      hinge = 'quadratic', 
      options = list(decomposition= 'chol'), 
      mc.cores = 1)
})
    
test_that(
  desc = "Only kernel changes in crossvalidation", {
    model <- svmmajcrossval(
      X.base, y.base, verbose = TRUE, 
      search.grid = list(kernel.sigma = 2^seq(-1,1), lambda = 10^seq(-2,2)),
      hinge = 'quadratic', 
      options = list(decomposition= 'chol'),
      kernel = rbfdot,
      mc.cores = 2)
  })

## TODO
# - test for single grid.point (including plot?)