Graph-based Multi-view Binary Learning for image clustering

Abstract Hashing techniques, also known as binary code learning, have recently attracted increasing attention in large-scale data analysis and storage. Generally, most existing hash clustering methods are single-view ones, which lack complete structure or complementary information from multiple views. For tasks, hashing research mainly mapped the original into Hamming space while heavily ignoring feature structure. To solve these problems, we propose a novel algorithm for clustering, adopts graph embedding to preserve structure, called Graph-based Multi-view Binary Learning (GMBL). The learning combines of different views clustering. Specifically, GMBL preserves local linear relationship utilizing Laplacian matrix aim maintain graph-based space. Moreover, by automatically assigning weights each view improve performance, takes distinctive contributions considerations. Besides, an alternating iterative optimization method is designed resulting problems. Extensive experimental results on five public datasets provided reveal effectiveness its superior performance over other state-of-the-art alternatives.