摘要
文章提出了一种基于投影梯度法的非负矩阵分解稀疏算法,该算法通过引入基于投影梯度的迭代方法,来解决加向量1-范数约束以及加向量2-范数约束的非负矩阵分解问题,得到了局部最优解。通过实验表明该算法在分解时间以及基矩阵的稀疏度表达能力上优于NMF算法和SNMF算法。
A sparse algorithm for non-negative matrix factorization based on projection gradient method is proposed.This algorithm introduces iteration method based on projection gradient to solve the issue of non-negative matrix factorization,which inclueds vector 1-norm constraint and vector 2-norm constraint,getting the local optimal solution.This algorithm is superior to both NMF algorithm and SNMF algorithm at the factorization time aspect and sparse degree ability of base matrix aspect,which is proved in the experiments.
出处
《计算机与数字工程》
2012年第12期20-22,59,共4页
Computer & Digital Engineering
关键词
非负矩阵分解
投影梯度法
稀疏算法
non-negative matrix decomposition
projection gradient method
sparse algorithm
