Incomplete multiview clustering (IMC) has attracted considerable attention as it can flexibly fuse the multiview information when part of the view samples are unobserved. Considering that the main challenge of IMC is the unobserved samples causing the information loss, in this article, we propose a novel IMC model to complete the unobserved samples, named cross-view feature transformation-based incomplete multiview clustering (CFTIMC). Unlike the previous completion methods that simply complete the unobserved samples with the average value or involve a third subspace, CFTIMC directly models the alignment relation between views by the cross-view feature transformation for completion, where only the overlapping views and its overlapping parts are selected for avoiding the affection of the nonoverlapping parts. Simultaneously, this completion process is integrated into multiview graph learning, where they can boost each other. To solve the proposed model, we develop an alternate optimization scheme-based iterative algorithm, together with its computational cost analysis and convergence analysis. Finally, the clustering experiments on several benchmark datasets verify that our method can outperform the related state-of-the-art methods.

Graph-based semi-supervised learning (GSSL) is a quite important technology due to its effectiveness in practice. Existing GSSL works often treat the given labels equally and ignore the unbalance importance of labels. In some inaccurate systems, the collected labels usually contain noise (noisy labels) and the methods treating labels equally suffer from the label noise. In this article, we propose a novel label-weighted learning method on graph for semi-supervised classification under label noise, which allows considering the contribution differences of labels. In particular, the label dependency of data is revealed by graph constraints. With the help of this label dependency, the proposed method develops the strategy of adaptive label weight, where label weights are assigned to labels adaptively. Accordingly, an efficient algorithm is developed to solve the proposed optimization objective, where each subproblem has a closed-form solution. Experimental results on a synthetic dataset and several real-world datasets show the advantage of the proposed method, compared to the state-of-the-art methods.