Abstract: When existing Non-negative Matrix Factorization(NMF) algorithm is applied to a problem of incremental scale, the consumption of space and time behaves inefficiency. This paper proposes an Incremental Nonnegative Matrix Factorization(INMF) algorithm, which uses partitioned matrix theory to reduce the computing scale, and uses decomposition results already derived to avoid re-calculating every time. Experimental results show that the algorithm performs efficiently for saving computing resources.

Key words: Non-negative Matrix Factorization(NMF), matrix factorization, incremental algorithm

摘要： 针对现有的非负矩阵分解算法在应用于问题规模逐渐增大的情形时，运算规模随之增大、空间和时间效率不高的情况，提出一种增量式非负矩阵分解算法，使用分块矩阵的思想降低运算规模，利用上一步的分解结果参与运算从而避免重复运算。实验结果表明，该算法对节约计算资源是有效的。

关键词: 非负矩阵分解, 矩阵分解, 增量式算法

CLC Number: 

• TP311.12