Deep Semi-Supervised Learning via Dynamic Anchor Graph Embedding Learning

Zihao Wang, Enmei Tu, Zhicheng Lee

This paper proposed a novel deep semi-supervised dynamic anchor graph embedding learning algorithm. The method has shown promise in text and image recognition tasks, especially when labeled data is limited. By introducing an auxiliary unsupervised task of predicting the neighborhood context in the graph, the approach effectively mines the structure information provided by abundant unlabeled data.