An overview of proxy-label approaches for semi-supervised learning
My Take
The relative expense and unavailability of labelled datasets is a major detractor from the utility of supervised learning techniques. On the other hand, unsupervised learning has a high degree of uncertainty, which brings its own pitfalls. I’ve also been interested to see if there are ways to bridge the gap, without just overfitting wholesale.
Their Take
Unsupervised learning constitutes one of the main challenges for current machine learning models and one of the key elements that is missing for general artificial intelligence. While unsupervised learning on its own is still elusive, researchers have a made a lot of progress in combining unsupervised learning with supervised learning. This branch of machine learning research is called semi-supervised learning.
http://ruder.io/semi-supervised/