科研之友
微信
新浪微博
Facebook
分享链接
摘要
In this paper, a novel regularized nonnegative matrix factorization (NMF) method, called neighbors isometric embedding nonnegative matrix factorization (NINMF), is proposed. The key idea of the NINMF method is to incorporate neighbors isometric regularized constraint in the optimization of the NMF. Hence, the NINMF is able to extract the representation space that preserves neighbor isometric geometry structure. Like most of the graph regularized NMFs, the NINMF method also finds a similarity weights matrix. However, the difference of our proposed method is that the NINMF simultaneously builds similarity weight matrix and performs data representation. The proposed method was applied to solve the problem of image representation using the well-known ORL, Yale and extended YaleB image data sets. The experimental results demonstrate the effectiveness of the proposed NINMF method for image representation.
