Multilabel

[luizfacury]

Description

This pull request adds support for multi-label classification to the OnlineKNN callback, along with related validation, warnings, and tests. The main changes include new arguments for multi-label mode, logic to handle multi-label voting in KNN prediction, and additional warnings for AUROC metric usage. Unit tests have also been added to ensure correct initialization and computation for multi-label scenarios.

Multi-label KNN support:

• Added multi_label and num_classes arguments to the OnlineKNN class, with validation to require num_classes when multi_label=True. [1] [2] [3]
• Implemented multi-label voting logic in _compute_knn_predictions, allowing correct prediction computation for multi-label datasets.

Metric handling and warnings:

• Added a warning in both OnlineKNN and probe callbacks when AUROC metrics are detected, informing users about required target label types for torchmetrics. [1] [2]

Testing:

• Added unit tests to verify multi-label initialization validation, AUROC warnings, and correct multi-label KNN voting math.

Checklist

• I have read the Contributing document.
• The documentation is up-to-date with the changes I made (check build artifacts).
• All tests passed, and additional code has been covered with new tests.
• [] I have added the PR to the RELEASES.rst