基于鲁棒非负矩阵分解的增量式学习研究  

作　　者：杨亮东 赵妍杰 李亚东 

机构地区：[1]兰州资源环境职业技术大学,兰州730021 [2]兰州博文科技学院,兰州730101

出　　处：《科技创新与应用》2025年第8期64-67,共4页Technology Innovation and Application

基　　金：兰州资源环境职业技术大学校级科技项目(X2023A-05)。

摘　　要：针对随着新增样本增多鲁棒非负矩阵分解的运算规模不断增大的现象,提出一种鲁棒非负矩阵分解的增量式学习算法。该算法对初始和新增样本采用L2,1范数来度量,首先对初始样本进行鲁棒非负矩阵分解;然后利用其分解结果参与后续迭代运算。通过对ORL和YALE人脸数据库进行实验得出,与鲁棒非负矩阵分解算法、稀疏限制的非负矩阵分解算法结果进行对比,该算法的目标函数值在求解时最先达到平衡,得到最佳收敛效果节省运算时间。Aiming at the phenomenon that the operation scale of robust non-negative matrix factorization continues to increase with the increase of new samples,an incremental learning algorithm for robust non-negative matrix factorization is proposed.The algorithm uses the L2,1 norm to measure the initial and new samples.First,the initial samples are subjected to robust non-negative matrix decomposition,and then the decomposition results are used to participate in subsequent iterative operations.Experiments on ORL and YALE face databases show that compared with the results of the robust non-negative matrix decomposition algorithm and the sparsity restricted non-negative matrix decomposition algorithm,the objective function value of this algorithm reaches balance first when solving,and the best convergence effect is obtained.Save computing time.

关 键 词：增量式学习 鲁棒非负矩阵分解 新增样本 稀疏限制 L2 1范数 

分 类 号：O151.21[理学—数学]