Multi-label classifiers and recommendation for prune method

[avostryakov]

Thanks for cleanlab!

Do you have examples of multi-label classification tasks? I'm curious what prune method to use in this case: prune by noise rate, prune by class, or both?

[sylvainma]

I am also looking for an example when the target y is a multi-output one:

[[0 1 0 0 1], 
 [1 0 0 0 1], 
 ...]

Any trick?

[jwmueller]

In case there is any chance this information is still useful to you:

In cleanlab v2.2, we did a major overhaul of the functionality for multi-label classification. cleanlab's new label error detection capabilities for multi-label classification are much more powerful and easy to use!

Get started here: https://docs.cleanlab.ai/stable/tutorials/multilabel_classification.html

This tutorial demonstrates how to handle multi-output targets in the format you have.

We also published a paper detailing the new algorithms cleanlab v2.2 uses to estimate label errors in multi-label datasets:
Identifying Incorrect Annotations in Multi-Label Classification Data

Through extensive benchmarking, this paper answers questions about which label error detection algorithms to use for multi-label data (that are the cleanlab default settings post v2.2).

And definitely please open a new issue (stating which cleanlab version you are using) if you still encounter any problems using cleanlab for multi-label datasets!