Multi-View Learning over Structured and Non-Identical Outputs

In many machine learning problems, labeled training data is limited but unlabeled data is ample. Some of these problems have instances that can be factored into multiple views, each of which is nea