Bases: object
Variance ratio for classification/multiple choice.
Measures disagreement in predictions across samples.
Defined as: VR = 1 - (most_common_count / total_samples)
-
__init__()
Methods
-
static compute(predictions)[source]
Compute variance ratio.
- Parameters:
predictions (List[int]) – List of predicted class indices
- Return type:
float
- Returns:
Variance ratio (0-1), higher = more uncertainty
