Consensus graph and spectral representation for one-step multi-view kernel based clustering

Recently, multi-view clustering has received much attention in the fields of machine learning and pattern recognition. Spectral for single multiple views been common solution. Despite its good performance, it a major limitation: requires an extra step clustering. This step, which could be famous k-means clustering, depends heavily on initialization, may affect quality result. To overcome this problem, new method called Multi-view Clustering via Consensus Graph Learning Nonnegative Embedding (MVCGE) is presented paper. In proposed approach, consensus affinity matrix (graph matrix), representation cluster index (nonnegative embedding) are learned simultaneously unified framework. Our takes as input different kernel matrices corresponding to views. The model integrates two interesting constraints: (i) indices should smooth possible over graph (ii) set close convolution representation. no post-processing such or spectral rotation required. approach tested with real synthetic datasets. experiments performed show that performs well compared many state-of-the-art approaches.