Multi-view learning arouses vast amount of interest in the past decades with numerous real-world applications in web page analysis, bioinformatics, image processing and so on. Unlike the most previous works following the idea of co-training, in this