dataeval.core.rank_result_stratified¶
-
dataeval.core.rank_result_stratified(result, num_bins=
50)¶ Transform RankResult indices using stratified sampling.
Takes a RankResult and applies stratified sampling to balance selection across score bins.
- Parameters:¶
- result : RankResult¶
Ranking result including indices and scores.
- num_bins : int, default 50¶
Number of bins for stratification.
- Returns:¶
Reordered indices in easy_first order with stratified sampling applied.
- Return type:¶
NDArray[np.intp]
- Raises:¶
ValueError – If result does not contain scores.