Robust nonnegative matrix factorization with structure regularization

Huang, Qi; Yin, Xuesong; Chen, Songcan; Wang, Yigang; Chen, Bowen

doi:10.1016/j.neucom.2020.06.049

Cited by 30 publications

(3 citation statements)

References 60 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The performance of ONMF is better than that of L21NMF in six data sets and NMF in all data sets, which validates that orthogonal constraint in NMF performs well for clustering tasks [3]. (7) It is worth noting that NMF performs almost as well as manifold learning-based algorithms on the Face94 and TOX_171 data sets, but it is much worse than RGNMF, especially on the TOX_171 data set. This indicates that is not enough to improve the performance of NMF only by encoding the local structure, but also by considering other information, such as the global structure.…”

Section: Comparison Of Clustering Performancesupporting

confidence: 60%

“…In addition, we use a big data set MNIST 4 to evaluate the performance of all algorithms. The data contained in the above-mentioned data sets are real-world and used by the state-of-the-art algorithms [7,23,28,41,46,66,67]. ORL and Georgia are two relatively new data sets.…”

Section: A Data Setsmentioning

confidence: 99%

“…As a popular technique for data representation, nonnegative matrix factorization (NMF) has been successfully applied in computational intelligence [1,2], machine learning [3,4] and data mining [5,6,7]. Its power lies in its ability to give meaningful decompositions of data into two nonnegative matrices.…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Robust Graph Regularized Nonnegative Matrix Factorization

et al. 2022

Self Cite

View full text Add to dashboard Cite

Nonnegative Matrix Factorization (NMF) has become a popular technique for dimensionality reduction, and been widely used in machine learning, computer vision, and data mining. Existing unsupervised NMF methods impose the intrinsic geometric constraint on the encoding matrix, which only indirectly affects the base matrix. Moreover, they ignore the global structure of the data space. To address these issues, in this paper we propose a novel unsupervised NMF learning framework, called Robust Graph regularized Nonnegative Matrix Factorization (RGNMF). RGNMF constructs a sparse graph imposed on the basis matrix to catch the global structure and preserve the discriminative information. And it models the local structure by building a k-NN graph constrained on the encoding matrix, which gains the compact representation. Consequently, RGNMF not only respects the global structure, but also depicts the local structure. In addition, it employs such a L2,1-norm cost function to decompose the basis matrix and encoding matrix that its robustness can be improved. Further, it imposes the L2,1-norm constraint on the basis matrix to choose the discriminative feature. Hence, RGNMF can gain the robust discriminative representation by combining structure learning and L2,1-norm constraints imposed on the basis matrix and encoding matrix. Extensive experiments on real-world problems demonstrate that RGNMF achieves better clustering results than the state-of-the-art approaches.

show abstract

Section: Comparison Of Clustering Performancesupporting

confidence: 60%