Researchers at Ryerson University in Canada have recently published their work on Deep Super Learning. In short, super learning is an ensemble meant to find the best mix of learning algorithms:
"The deep super learner is flexible, adaptable, and easy to train with good performance across different tasks using identical hyper-parameter values" [source]
You can read more about the technical aspects of this in the release paper. It didn't take long until someone made an implementation of this concept in scikit-learn.
This repo is still in its infancy, so don't expect too much documentation just yet. But I'd suspect that if it catches and if it finds wide implementation within the field, documentation is not going to be a problem. And since we're talking about github and open-source, anyone can contribute to its development.
To stay in touch with me, follow 
Cristi Vlad Self-Experimenter and Author