Skip to contents

Evaluate fitness with two competing objectives

Source: R/RcppExports.R
evaluate_fitness_cpp.Rd

Objectives are always minimised. Weighted columns are returned so the caller can inspect each objective (weights are recycled when shorter than the number of objectives).

Usage

evaluate_fitness_cpp(
  genomic_data,
  population,
  weights,
  labels = NULL,
  representation = "real"
)

Arguments

genomic_data

Numeric matrix of genomic data (rows: genes, columns: samples).

population

Numeric matrix of individuals (rows: individuals, columns: genes).

weights

Numeric vector of objective weights (recycled if shorter than two).

labels

Optional class labels (length == number of samples); an integer, numeric, factor, character or logical vector is accepted and mapped to integer class codes. Required for representation = "binary".

representation

Either "real" or "binary".

Value

Numeric matrix of fitness scores (rows: individuals, columns: objectives).

Details

In "real" mode individuals are continuous expression profiles:

  • f1: squared Euclidean distance between the individual and every observed sample (data fidelity).

  • f2: mean absolute gene value (L1 parsimony).

In "binary" mode individuals are gene-selection indicators and labels (one class label per sample) are required:

  • f1: negated mean ANOVA F-statistic of the selected genes (predictive signal).

  • f2: fraction of selected genes (cardinality / sparsity).

See also

initialize_population_cpp(), selection_cpp(), crossover_cpp(), mutation_cpp(), replacement_cpp(), bioga_main_cpp()

Examples

genomic_data <- matrix(rnorm(100), nrow = 10, ncol = 10)
population <- BioGA::initialize_population_cpp(genomic_data, population_size = 5)
BioGA::evaluate_fitness_cpp(genomic_data, population, weights = c(1.0, 0.5))
#>          [,1]      [,2]
#> [1,] 211.7487 0.5084159
#> [2,] 171.4019 0.3824670
#> [3,] 190.9419 0.3457173
#> [4,] 212.2748 0.4887263
#> [5,] 228.8821 0.4212201