Run a detector against positives and generated negatives — build_autotype

Scores detector on the real positives P and on negatives generated at each mutation tier, returning one tidy table. Used to build the human-readable artifacts in data-types/.

Usage

build_autotype_results(
  detector,
  P,
  tiers = c("S1", "S2", "S3", "SP"),
  n_per_positive = 20,
  p = 0.5,
  seed = 514,
  threshold = 0.5
)

Arguments

detector: A function taking a character vector and returning a logical (or numeric score) vector.
P: A character vector of positive examples.
tiers: Mutation tiers to include.
n_per_positive: Mutants attempted per positive per tier.
p: Per-character mutation probability.
seed: Integer seed.
threshold: Score threshold used when detector returns numeric.

Value

A data frame with columns case, type ("valid" or a tier), f_match (raw detector output), and f_predict (logical prediction).