[scikit-learn] MLP Maxout activation proposal

[深谷亮祐]

Hi everyone,

My name is Ryosuke Fukatani.
I'm joining a scikit-learn community and really excited to work with all.

Today I propose new feature "maxout activation" for MLP.
Maxout activation achived high performance
classification.(http://www.jmlr.org/proceedings/papers/v28/goodfellow13.pdf).

And it seemed that only a few layers are needed to use maxout.
I think it is suitable feature for light weight MLP by scikit learn.

If it's OK, I'd work on it.
Best regards