An incomplete multi-view clustering approach using subspace alignment constraint

For partially missing multi-view data, existing incomplete multi-view clustering approaches suffer from the effects of misplaced neighbour relationships and changes in subspace structure caused by data incompleteness and imbalances in clustering information among views. In this paper, an incomplete multi-view clustering approach is proposed using subspace alignment constraints. Firstly, subspace alignment constraint is proposed among incomplete subspace representations to mitigate the misalignment-related changes in clustering structures. An adaptive weight is proposed to balance clustering information among views. Secondly, an incomplete multi-view clustering model is proposed by integrating subspace alignment constraint and inter-view adaptive representation, and a clustering algorithm is designed using an alternate minimizing optimization strategy. In the end, experimental results on benchmark datasets validate the algorithm’s excellent clustering performance for incomplete multi-view datasets.