Regularized Multi-view Spectral Clustering With Noisy Data

Multi-view clustering has attracted much attention in recent years. When the data is noisy, traditional multi-view algorithm might not receive good results. An entropy perturbed multi-view spectral clustering method is proposed to solve this problem. In order to reduce the impact of noise on the data, entropy is used as a measure to perturb the Laplacian matrix within a threshold, trying to find a set of eigenvectors carrying largest information. Canonical angle is used to measure the difference in clustering ability between each view and the consensus view to assign weights for views. Experimental results show that the proposed method has good performance on both noisy and non-noisy datasets.