|
| 摘要: |
| 本文研究了基于核技巧的L2，1范数非负矩阵分解在图像聚类中的问题.利用基于核的稀疏鲁棒非负矩阵分解方法，获得了算法良好的稀疏性和鲁棒性，提高了聚类性能，该方法也可以推广到文本聚类的应用. |
| 关键词: 非负矩阵分解 核技巧 L2,1范数 稀疏性 鲁棒性 |
| DOI： |
| 分类号:O235 |
| 基金项目:国家自然科学基金（11601012；71561008）；广西自然科学基金（2018GXNSFAA138169）；广西密码学与信息安全重点实验室研究课题（GCIS201708）；广西自动检测技术与仪器重点实验室基金（YQ16112；YQ18112）；宁夏自然科学基金（NZ17103）；桂林电子科技大学研究生优秀学位论文培育项目资助（16YJPYSS22）. |
|
| KERNEL-BASED L2,1 NORM NON-NEGATIVE MATRIX FACTORIZATION IN IMAGE CLUSTERING |
|
YU Jiang-lan,LI Xiang-li,DONG Xiao-liang
|
| Abstract: |
| The problem of norm non-negative matrix factorization with L2,1 is studied based on kernel technique in image clustering. By kernel-based sparse robust non-negative matrix factorization method, the sparseness and robustness of the algorithm are obtained, and the clustering performance is improved. This method can also be extended to the application of text clustering. |
| Key words: non-negative matrix factorization kernel trick L2,1 norm sparsity robustness |
