A novel regularized asymmetric non-negative matrix factorization for text clustering
作者：
Highlights：
• NMF generates low-rank data while keeping the non-negativity of elements.
• Multiplicative updating rules in NMF warrant the convergence of the algorithm.
• Regularized NMF imposes constraints as regularization variables.
• The proposed RANMF uses Kullback–Leibler divergence as the cost function.
• RANMF applies penalties on factorized matrices to create a generic algorithm.
摘要
•NMF generates low-rank data while keeping the non-negativity of elements.•Multiplicative updating rules in NMF warrant the convergence of the algorithm.•Regularized NMF imposes constraints as regularization variables.•The proposed RANMF uses Kullback–Leibler divergence as the cost function.•RANMF applies penalties on factorized matrices to create a generic algorithm.
论文关键词：Text clustering,Semantic features,Non-negative matrix factorization,Multiplicative updating rules
论文评审过程：Received 26 March 2021, Revised 30 May 2021, Accepted 12 July 2021, Available online 28 July 2021, Version of Record 28 July 2021.
论文官网地址：https://doi.org/10.1016/j.ipm.2021.102694
