Skip to content

Pre-defined Pair Classification

Same as all-to-all pair classification, but operates on datasets with pre-defined pairs (e.g., authorship verification). Episode size and episodes-per-class are set automatically.

Metrics

  • EER (Equal Error Rate): lower is better
  • AUC: higher is better
  • AUC@FPR: AUC at false positive rate thresholds 0.01, 0.05, 0.10, 0.20, 0.30, 0.50

CLI

steb pre_defined_pair_classification rrivera1849/LUAR-MUD --dataset pan15_authorship_verification_english_test

Datasets

List available datasets:

steb pre_defined_pair_classification --list-datasets