Add a multi-label classification task

[raghavrv]

Add a dataset with multi-label data.

Refer #5105

Preferably the emotions dataset

[ghost]

I would like to solve this issue. Cause I have just achieved a multi-label classification using Matlab, I can try to collect data from the emotions dataset and test it using sklearn.

[raghavrv]

@wujw13 Please go ahead!!

[ghost]

@rvraghav93 I think the emotions dataset may not be used. In its End User License Agreement, it suggests that "The user may not distribute the dataset or portions thereof in any way". Is there second choice of a dataset that fits requirements?

[raghavrv]

Wait lets ping @arjoly or @amueller for suggestions :)

[arjoly]

An another good choice is the "scene" dataset. I hope it is not "too" big in term of memory footprint.

[ghost]

All right. I will check this dataset.

[amueller]

yeast is smaller, right?

[arjoly]

yes it's a bit smaller. yeast can be easily obtain using mldata (through fetch_mldata).

[kshitij10496]

I would like to work on this issue if no one is working on it

[kshitij10496]

Ping @rvraghav93 @arjoly @amueller I just want to make sure whether we are going to add the yeast dataset or something else ?

[raghavrv]

@kshitij10496 Yes please go ahead and add the yeast dataset. Refer #5325

[kshitij10496]

@rvraghav93 Thanks I am working on it

[gxyd]

@amueller as you said in your comment, is that still valid?

Will the first step towards completing PR https://github.com/scikit-learn/scikit-learn/pull/5960 be to address your comment?

[gxyd]

Also I don't see no 'yeast' dataset included with scikit-learn.

[jnothman]

Don't we have rcv1 as a real multilabel dataset with a fetcher? Yes, I think this can be closed.