Manifold Coarse Graining for Online Semi-Supervised Learning

Since labeling data is often both laborious and costly, the labeled data available in many applications is rather limited. Active learning is a learning approach which actively selects unlabeled data