PHASL-NMF: Hierarchical ALS Based Power Non-Negative Matrix Factorization
Yuan Luo, Bing Han, Nian Zhang, Peng Zhou, Jiang Xiong, Yuzhi Zhao, Li Chen, Xiangguang Dai
Abstract
Yuan Luo, Bing Han, Nian Zhang, Peng Zhou, Jiang Xiong, Yuzhi Zhao, Li Chen, Xiangguang Dai
Abstract
The non-negative matrix factorization (NMF) has been found an effective clustering algorithm and it outperforms the classical k-means algorithm. Existing researches mainly focus on the problem of reducing the decomposition error between two decomposition matrices and the original matrix. In this paper, inspired by the power k-means algorithm and the hierarchical alternating least square NMF, we propose a novel NMF algorithm called power NMF (PHALS-NMF), which introduces the power mean to reduce decomposition error. Massive experiments on several datasets show the feasibility and effectiveness of PHALS-NMF.
A significance statement is not available in the OpenAlex record.
A contribution statement is not available in the OpenAlex record.
Method details are not available in the OpenAlex metadata.
Findings are not separately available in the OpenAlex metadata.
Limitations are not available in the OpenAlex metadata.
Application details are not available in the OpenAlex metadata.
The non-negative matrix factorization (NMF) has been found an effective clustering algorithm and it outperforms the classical k-means algorithm. Existing researches mainly focus on the problem of reducing the decomposition error between two decomposition matrices and the original matrix. In this paper, inspired by the power k-means algorithm and the hierarchical alternating least square NMF, we propose a novel NMF algorithm called power NMF (PHALS-NMF), which introduces the power mean to reduce decomposition error. Massive experiments on several datasets show the feasibility and effectiveness of PHALS-NMF.
Key concepts: Non-negative matrix factorization, Matrix decomposition, Computer science, Cluster analysis, Decomposition, Focus (optics), Factorization, Matrix (chemical analysis)