inst/tinytest/test-c-implementation.R

In anticlust: Subset Partitioning via Anticlustering

library("anticlust")

set.seed(123)
N <- sample(40:80, size = 1)
M <- sample(1:5, size = 1)
K <- sample(2:5, size = 1)
features <- matrix(rnorm(N * M), ncol = M)
clusters <- anticlust:::initialize_clusters(N, K, NULL)
cl1 <- anticlustering(features, clusters, objective = "variance")
cl2 <- anticlustering(features, clusters, variance_objective)
expect_true(all(cl1 == cl2))

# now also use a categorical constraint
set.seed(71207)
C <- sample(2:5, size = 1)
categories <- sample(C, size = N, replace = TRUE)
clusters <- anticlust:::initialize_clusters(N, K, categories)
cl1 <- anticlustering(
  features, 
  clusters, 
  objective = "variance", 
  categories = categories
)
cl2 <- anticlustering(
  features, 
  clusters, 
  objective = variance_objective, 
  categories = categories
)
expect_true(all(cl1 == cl2))

# Include a test for the edge cases (i.e., categories with 1 or N members)
categories[1] <- C + 1 
clusters <- anticlust:::initialize_clusters(N, K, categories)
cl1 <- anticlustering(
  features, 
  clusters, 
  objective = "variance", 
  categories = categories
)
cl2 <- anticlustering(
  features, 
  clusters, 
  objective = variance_objective, 
  categories = categories
)
expect_true(all(cl1 == cl2))

# Include a test for the edge cases (i.e., categories with 1 or N members)
categories <- rep(1, N)
clusters <- anticlust:::initialize_clusters(N, K, categories)
cl1 <- anticlustering(
  features, 
  clusters, 
  objective = "variance", 
  categories = categories
)
cl2 <- anticlustering(
  features, 
  clusters, 
  objective = variance_objective, 
  categories = categories
)
expect_true(all(cl1 == cl2))

# Use diversity objective
set.seed(123)
clusters <- anticlust:::initialize_clusters(N, K, NULL)
cl1 <- anticlustering(features, clusters, objective = "diversity")
cl2 <- anticlustering(features, clusters, diversity_objective)
expect_true(all(cl1 == cl2))

# now also use a categorical constraint
set.seed(71207)
C <- sample(2:5, size = 1)
categories <- sample(C, size = N, replace = TRUE)
clusters <- anticlust:::initialize_clusters(N, K, categories)
cl1 <- anticlustering(
  features, 
  clusters, 
  objective = "diversity", 
  categories = categories
)
cl2 <- anticlustering(
  features, 
  clusters, 
  objective = diversity_objective, 
  categories = categories
)
expect_true(all(cl1 == cl2))

# now also use preclustering
set.seed(71207)
C <- sample(2:5, size = 1)
categories <- sample(C, size = N, replace = TRUE)
clusters <- anticlust:::initialize_clusters(N, K, categories)
cl1 <- anticlustering(
  features, 
  clusters, 
  objective = "diversity", 
  categories = categories
)
cl2 <- anticlustering(
  features, 
  clusters, 
  objective = diversity_objective, 
  categories = categories
)
expect_true(all(cl1 == cl2))

# test dispersion objective
set.seed(123)
N <- 50 # number of elements
M <- 2  # number of variables per element
K <- 2  # number of clusters
random_data <- matrix(rnorm(N * M), ncol = M)
random_clusters <- sample(rep_len(1:K, N))

# Maximize the dispersion 
optimized_clusters <- anticlustering(
  random_data,
  K = random_clusters, 
  objective = dispersion_objective
)
dispersion_objective(random_data, optimized_clusters)

optimized_clusters2 <- anticlustering(
  random_data,
  K = random_clusters, 
  objective = "dispersion"
)
expect_true(all(optimized_clusters == optimized_clusters2))