I see the module is able to predict multi-class but is it able to predict multi-label?

[vmgustavo]

In the example below, there are observations labeled as classes 0, 1, 2 or 3. If there is a DAG defining the hierarchy for the classes then it is possible to train a model using hiclass.

obs_id	class
6653	2
1241	1
5123	0
3312	1
4421	2
6316	1
7281	1
6345	0
8397	3

But what if the case is as follows:

obs_id	class0	class1	class2	class3
6653	0	0	1	0
1241	1	1	0	1
5123	0	0	1	0
3312	0	0	0	1
4421	0	0	1	1
6316	1	0	1	0
7281	0	1	0	1
6345	1	1	0	1

I have multiple binary labels that represent each observation and those labels can be organized in a DAG structure to define the hierarchy. Is it possible to train a model using hiclass?

From the documentation I am assuming it is not possible. Is the current framework able to expand to this scenario?

[dkoppes]

I wonder this as well, in previous releases I did see documentation about multi-label classification, but now it seems to be removed.