非负矩阵分解的分布式算法

A distributed learning algorithm for nonnegative matrix factorization

提出一种解决大规模非负矩阵分解的分布式算法.非负矩阵分解一直是矩阵分解领域中的热点问题之一,已有一些相关的算法.但是,对于大规模的非负矩阵,至今尚无高效的方法.本文采用近来解决大数据的分布式思想和并行式计算方法,并将它们与传统的矩阵分解算法相结合,提出一种基于并行式计算的分布式网络算法,以此实现大规模的非负矩阵分解问题.实验结果表明,所提出的算法较一般的分布式算法与集中式矩阵分解的算法更加有效和快速.

A distributed learning algorithm is put forward for dissolving the factorization of large-scale nonnegative matrixes.. The factorization of nonnegative matrixes is a hot problem in this field with many effective algorithms. However, the large-scale nonnegative matrixes, there have not been any highly valid algorithms. We combined the distributed concept and the parallel computing with the traditional matrix factorization methods to develop a distributed learning algorithm for complex nonnegative matrix factorization. The simulation experiments show that the proposed algorithm is more efficient and faster than the traditional distributed learning algorithm and matrix factorization methods.