Graph Convolutional Network Combined with Semantic Feature Guidance for Deep Clustering

The performances of semisupervised clustering for unlabeled data are often superior to those unsupervised learning, which indicates that semantic information attached clusters can significantly improve feature representation capability. In a graph convolutional network (GCN), each node contains about itself and its neighbors is beneficial common unique features among samples. Combining these findings, we propose deep method based on GCN guidance (GFDC) in used as generator, with softmax layer performs assignment. First, the diversity amount input enhanced generate highly useful representations downstream tasks. Subsequently, topological constructed express spatial relationship features. For pair datasets, correspondence constraints regularize loss, outputs iteratively optimized. Three external evaluation indicators, i.e., accuracy, normalized mutual information, adjusted Rand index, an internal indicator, Davidson-Bouldin index (DBI), employed evaluate performances. Experimental results eight public datasets show GFDC algorithm better than majority competitive methods, accuracy 20% higher best United States Postal Service dataset. also has highest smaller Amazon Caltech datasets. Moreover, DBI dispersion cluster distribution compactness within cluster.