摘要
Nonnegative matrix factorization (NMF) was a classic model for dimensional reduction. Manhattan NMF is a variant version of NMF that uses a L-1-norm cost function as the objective function instead of the L-2-norm cost function. Manhattan NMF can be formulated as a nonconvex nonsmooth optimization problem. An algorithm framework for solving the Manhattan NMF problem based on the alternating direction method of multiplication is presented to us. Compared with the existed algorithm, our proposed algorithm is more effective by experiments on synthetic and real data sets.
学科与关键词
关键词
Nonnegative matrix factorizationL-1-normnonconvex nonsmooth
- Computing a nonnegative matrix factorization -- provablySanjeev, Arora; Rong, Ge; Ravindran, Kannan; Ankur, Moitrathe 44th symposium, 2012-05-19 To 2012-05-22.被引 60
- A survey of deep nonnegative matrix factorizationChen, Wen-Sheng; Zeng, Qianwen; Pan, Binbin*NEUROCOMPUTING, 2022, 491: 305-320.被引 29
- 稀疏性正则化非负矩阵分解的在线学习方法薛模根; 徐国明; 王峰模式识别与人工智能, 2013, (03): 242-246.
