Fast and General Incomplete Multi-view Adaptive Clustering

With the development of data collection technologies, multi-view clustering (MVC) has become an emerging research topic. The traditional MVC method cannot process incomplete views. In recent years, although many incomplete multi-view clustering methods have been proposed by many researchers, these methods still suffer from some limitations. For example, these methods all have parameters that need to be adjusted, or have high computational complexity and are not suitable for processing large-scale data. To make matters worse, these methods are not suitable for cases where there are no paired samples among multiple views. The above limitations make existing methods difficult to apply in practice. This paper proposes a Fast and General Incomplete Multi-view Adaptive Clustering (FGPMAC) method. The FGPMAC adopts an adaptive neighbor assignment strategy to independently construct the similarity matrix of each view, thereby it can handle the cases where there are no paired samples among multiple views, and eliminating the necessary to adjust the parameters. Moreover, by adopting a non-iterative approach, FGPMAC has low computational complexity and is suitable for large-scale datasets. Results of experiments on multiple real datasets fully demonstrate the advantages of FGPMAC, such as simplicity, effectiveness and superiority.