Create synthetic similarity matrix for testing pseudolabeling — create_synthetic_similarity

Generates a synthetic sparse similarity matrix with known cluster structure. Used for testing and demonstrating the pseudolabeling functions.

Usage

create_synthetic_similarity_matrix(
  n_samples = 1000,
  n_clusters = 20,
  within_cluster_sim = 0.8,
  between_cluster_sim = 0.1,
  sparsity = 0.1,
  noise_level = 0.1
)

Arguments

n_samples: Total number of samples
n_clusters: Number of true clusters
within_cluster_sim: Average similarity within clusters (default: 0.8)
between_cluster_sim: Average similarity between clusters (default: 0.1)
sparsity: Overall sparsity level (proportion of non-zero entries) (default: 0.1)
noise_level: Amount of noise to add to similarities (default: 0.1)

Value

A list containing:

sim_matrix: Sparse similarity matrix (dgCMatrix) with synthetic cluster structure
true_labels: Factor vector of true cluster assignments for validation
cluster_centers: Integer vector of cluster center indices

Examples

# \donttest{
# Create synthetic data
synthetic <- create_synthetic_similarity_matrix(n_samples = 500, 
                                                n_clusters = 10)

# Apply pseudolabeling
result <- assign_pseudolabels(synthetic$sim_matrix, verbose = TRUE)
#> Processing 500 samples with 9290 non-zero similarities
#> Adaptive threshold: 0.8682 ( 0.8 quantile of 9290 similarities)
#> 'as(<dsCMatrix>, "dgCMatrix")' is deprecated.
#> Use 'as(., "generalMatrix")' instead.
#> See help("Deprecated") and help("Matrix-deprecated").
#> After thresholding: 1858 connections remain
#> Found 17 initial clusters
#> Found 10 valid clusters after size filtering
#> Final result: 10 clusters, 493 samples assigned

# Compare with true labels
table(result$labels, synthetic$true_labels, useNA = "always")
#>             
#>              cluster_1 cluster_10 cluster_2 cluster_3 cluster_4 cluster_5
#>   anchor_001        49          0         0         0         0         0
#>   anchor_002         0          0        49         0         0         0
#>   anchor_003         0          0         0        50         0         0
#>   anchor_004         0          0         0         0        50         0
#>   anchor_005         0          0         0         0         0        48
#>   anchor_006         0          0         0         0         0         0
#>   anchor_007         0          0         0         0         0         0
#>   anchor_008         0          0         0         0         0         0
#>   anchor_009         0          0         0         0         0         0
#>   anchor_010         0         49         0         0         0         0
#>   <NA>               1          1         1         0         0         2
#>             
#>              cluster_6 cluster_7 cluster_8 cluster_9 <NA>
#>   anchor_001         0         0         0         0    0
#>   anchor_002         0         0         0         0    0
#>   anchor_003         0         0         0         0    0
#>   anchor_004         0         0         0         0    0
#>   anchor_005         0         0         0         0    0
#>   anchor_006        50         0         0         0    0
#>   anchor_007         0        50         0         0    0
#>   anchor_008         0         0        50         0    0
#>   anchor_009         0         0         0        48    0
#>   anchor_010         0         0         0         0    0
#>   <NA>               0         0         0         2    0
# }