inst/tinytest/test-ilp-setup.R

In anticlust: Subset Partitioning via Anticlustering

#test setup of ILP
library("anticlust")

#anticlustering_ilp
# ILP is set up as expected
conditions <- expand.grid(m = 1:4, p = 2:4)
for (k in 1:nrow(conditions)) {
  m_features <- conditions[k, "m"]
  p_anticlusters <- conditions[k, "p"]
  n_elements <- p_anticlusters * 2 # n must be multiplier of p
  features <- matrix(rnorm(n_elements * m_features), ncol = m_features)
  distances <- as.matrix(dist(features))
  ilp <- anticlust:::anticlustering_ilp(distances, p_anticlusters)
  expect_equal(nrow(ilp$constraints), choose(n_elements, 3) * 3 + n_elements)
  expect_equal(sum(is.na(ilp$constraints)), 0)
  
  ## Test that distance matrix is correctly transfered into objective vector
  distances <- as.matrix(distances)
  costs <- ilp$costs
  for (i in 1:n_elements) {
    for (j in 1:n_elements) {
      if (i >= j) next
      expect_equal(costs[costs$i == i & costs$j == j, ]$costs, distances[i, j])
    }
  }
}