CAJ | 학술논문

近年来,混合的分数矩阵范数l2,p(0< p≤1)在高维数据处理的特征选取中有很好的表现,其基本思想是利用了欧氏范数l2的光滑性和分数范数lp(0< p≤1)的稀疏性。大量实验数据表明,混合分数矩阵范数l2,p(0< p <1)不仅比传统的向量范数l1具有更好的联合稀疏性,对噪声的抗干扰性也更强。文中依据人脸数据的稀疏结构,建立基于混合矩阵范数l2,p(0< p≤1)极小化的特征选取模型,结合最近邻识别方法,提出了一类新的鲁棒人脸分类方法。在多个人脸数据集上的实验结果表明,基于分数矩阵范数的新模型比传统的人脸识别方法有更好的特征选择及分类效果。
근년래,혼합적분수구진범수l2,p(0< p≤1)재고유수거처리적특정선취중유흔호적표현,기기본사상시이용료구씨범수l2적광활성화분수범수lp(0< p≤1)적희소성。대량실험수거표명,혼합분수구진범수l2,p(0< p <1)불부비전통적향량범수l1구유경호적연합희소성,대조성적항간우성야경강。문중의거인검수거적희소결구,건립기우혼합구진범수l2,p(0< p≤1)겁소화적특정선취모형,결합최근린식별방법,제출료일류신적로봉인검분류방법。재다개인검수거집상적실험결과표명,기우분수구진범수적신모형비전통적인검식별방법유경호적특정선택급분류효과。
Recently,mixed fractional matrix norm l2,p(0 < p≤1) has exhibited good performance for feature selection in high-dimen-sional data processing. The natural idea is combining the smoothness of Euclid norm l2 and the sparsity of norm lp(0 < p≤1) . A variety of experimental results have showed that the fractional matrix norm l2,p(0 < p < 1) has better joint sparsity and stronger reliability than traditional vector norm l1 . In this paper,a l2,p(0 < p≤1) -based minimization is presented to select feature in face data. Combined with the nearest neighbor classification method,a robust face recognition method is proposed. The extensive experiments on face data sets have showed that the l2,p(0 < p < 1) -minimization model performs better than the state-of-the-art on feature selection and classification effect.