A network-based sparse and multi-manifold regularized multiple non-negative matrix factorization for multi-view clustering

• An approach based on NMF of multi-network for multi-view clustering is proposed. factorization with multiple regularizations proposed and developed. Extensive experimental studies were performed real data evaluation. Multi-view has attracted increasing attention in recent years since many sets are usually gathered from different sources or described by feature types. Amongst various existing algorithms, those that non-negative matrix (NMF) have exhibited superior performance. However, decomposing original directly fails to exploit global relationships between samples cannot be applied datasets not strictly non-negative. In this paper, a network-based sparse multi-manifold regularized (NSM_MNMF) proposed, where transformed into networks, used jointly factorize networks capturing the shared cluster structure embedded views. Furthermore, regularization incorporated keep intrinsic geometrical information network manifold space. Networks characterize intra-view similarity, joint reveals inter-view similarity across distinct views, while using decompose instead means NSM_MNMF can results interpretable. experiments conducted nine assess method illustrate outperforms other baseline approaches.