[MRG] Issue #111 - Handle multiclass/binary class target

[glemaitre]

No description provided.

[coveralls]

Coverage decreased (-1.3%) to 98.042% when pulling ff52bfc on glemaitre:issue_111 into 0c46346 on scikit-learn-contrib:master.

[glemaitre]

The decrease of the coverage come from disabling the unit test for the CNN method

[glemaitre]

@dvro @chkoar If you can have a look to the changes. It should be ready for a merging normally.

[chkoar]

In order to avoid code and docsting duplication I would define a new private class _BinarySamplerMixin that it will inherit SamplerMixin and I would change the fit there. Then all binary samplers should inherit from that class.

[glemaitre]

One or two classes: BinarySamplerMixin and MulticlassSamplerMixin?

[chkoar]

At first I thought to split the SamplerMixin in two classes like you said. But we want something temporarily just to pass the check_estimator test thats why I said a private class.

[coveralls]

Coverage decreased (-1.3%) to 98.005% when pulling e7ac93f on glemaitre:issue_111 into 0c46346 on scikit-learn-contrib:master.

[coveralls]

Coverage decreased (-1.3%) to 98.005% when pulling 60d8dbc on glemaitre:issue_111 into 0c46346 on scikit-learn-contrib:master.

[glemaitre]

@dvro @chkoar Good for another round

[chkoar]

Finally, will we follow auto-sklearn convention?
Wouldn't it be better to name it _estimator_properties?
Also, do we need to make get_properties public? I guess so.

[coveralls]

Coverage decreased (-1.4%) to 97.926% when pulling 012da67 on glemaitre:issue_111 into 0c46346 on scikit-learn-contrib:master.

[coveralls]

Coverage decreased (-1.4%) to 97.926% when pulling 012da67 on glemaitre:issue_111 into 0c46346 on scikit-learn-contrib:master.

[chkoar]

I would name it BaseBinarySampler

[glemaitre]

agree

[coveralls]

Coverage decreased (-1.3%) to 97.986% when pulling e9d8f1a on glemaitre:issue_111 into 0c46346 on scikit-learn-contrib:master.

[glemaitre]

@chkoar I made the changes