Incomplete Multi-View Clustering via Auto-Weighted Fusion in Partition Space

As a class of effective methods for incomplete multi-view clustering, graph-based algorithms have recently drawn wide attention. However, most them could use further improvement regarding the following aspects. First, in some models, all views are forced to share common similarity graph regardless severe consistency degeneration due views. Next, construction and cluster analysis sometimes performed separately. Finally, contribution difference individual is not always carefully considered. To address these issues simultaneously, this paper proposes an clustering algorithm based on auto-weighted fusion partition space. In our algorithm, information structure introduced into process learning construct desirable graph, space alleviate negative impact brought about by degradation, adaptively weighted reflect their different contributions tasks. subtasks collaboratively optimized united framework reach overall optimal result. Experimental results show that proposed method compares favorably with state-of-the-art methods.