Quadratic regularization projected Barzilai–Borwein method for nonnegative matrix factorization
Quadratic regularization projected Barzilai–Borwein method for nonnegative matrix factorization
复制标题
DOI：
10.1007/s10618-014-0390-x
复制
发表时间：
2015-11
影响因子：
4.8
通讯作者：
Yakui Huang;Hongwei Liu;Shuisheng Zhou
中科院分区：
文献类型：
--
作者：
Yakui Huang;Hongwei Liu;Shuisheng Zhou
本文基于交替非负最小二乘框架，提出了一种新的非负矩阵分解的有效方法，该方法使用二次正则化投影Barzilai-Borwein（QRPBB）方法求解子问题.在每次迭代中，QRPBB方法首先通过求解强凸二次最小化问题来生成一个点，该问题具有计算成本低的简单封闭形式解，然后应用投影Barzilai-Borwein方法来更新NMF的解。在较弱的条件下，得到了全局收敛性结果。在合成数据集和真实数据集上的数值比较表明，所提出的方法是有效的。
In this paper, based on the alternating nonnegative least squares framework, we present a new efficient method for nonnegative matrix factorization that uses a quadratic regularization projected Barzilai–Borwein (QRPBB) method to solve the subproblems. At each iteration, the QRPBB method first generates a point by solving a strongly convex quadratic minimization problem, which has a simple closed-form solution that is inexpensive to calculate, and then applies a projected Barzilai–Borwein method to update the solution of NMF. Global convergence result is established under mild conditions. Numerical comparisons of methods on both synthetic and real-world datasets show that the proposed method is efficient.