Classification
Hamming loss
Implemented
classification.hamming_lossDefinition
Fraction of labels predicted incorrectly (for single-label targets equal to 1 − accuracy).
Formula
(1/(n·L)) Σᵢ Σₗ 1[ŷᵢₗ ≠ yᵢₗ]
Range: [0, 1]
Inputs and outputs
- Y: multilabel indicator matrix
- P: predicted indicator matrix
Returns: MetricResult
Assumptions
No assumptions beyond valid, aligned inputs of the documented types.
Limitations
No metric-specific limitations are documented yet. Interpret the value alongside the task, data, and other metrics.
Python API
import evalsuite as es
es.hamming_loss(Y, P)References
- Tsoumakas G, Katakis I. Multi-label classification: an overview. Int J Data Warehousing and Mining. 2007;3(3):1-13.