Latent Multi-view Clustering Based Adaptive Graph Constraint

Graph-based multi-view clustering methods have demonstrated impressive outcomes in capturing the underlying manifold structure of data, leading to improved clustering performance. However, conventional graph-based methods overlook the significance of distinct features and rely solely on the learned graph based on raw features, potentially restraining their performance. To overcome these restrictions, we introduce an innovative method: Latent Multi-view Clustering Based Adaptive Graph Constraint (LMCAGC). Our method incorporates manifold information from the original data and utilizes adaptive learning graphs to capture the relationships among samples. Specifically, the initial high-dimensional data is employed to reconstruct the latent representation matrix, and we construct the global similarity matrix through a linear amalgamation of affinity matrices across all perspectives. Subsequently, the technique of manifold regularization is utilized to improve the performance of the latent presentation model. Our method combines latent representation and adaptive graph learning within a unified framework optimized via an alternating iteration algorithm. Exhaustive experiments conducted on eight benchmark data sets affirm the performance of the proposed method. The source code is available at: https: //github. com/Wenboli11/LMCAGC.