改进的非负矩阵因子分解算法在基因数据分析中的应用

Improved non-negative factorization in the analysis of gene expression data

提出一种改进的非负矩阵因子分解算法.在非负矩阵因子分解的迭代计算过程中加入了数据平滑处理来解决抖动问题,并用于一组白血病微阵列数据分析.实验结果表明,改进过的非负矩阵分解算法提高了分类的准确率,同时这个方法避免了NMF算法的"零值"问题.

Improvement non-negative matrix factorization （NMF） algorithm has been proposed. Data smoothing has been added in the iteration of the NMF algorithm to solve the dithering problem. The improved NMF algorithm is applied in the analysis of leukaemia microarray data. Experiment results show that the accuracy can be significantly improved with the proposed algorithm. Furthermore, the problem of ＇ zeros＇ for the traditional NMF algorithms can be easily tackled in our proposed method.