为了避免图像数据向量化后的维数灾难问题,以及增强对野值(outliers)及噪声的鲁棒性,该文提出一种基于L1-范数的2维线性判别分析(L1-norm-based Two-Dimensional Linear Discriminant Analysis,2DLDA-L1)降维方法。它充分利用L1-范数对...为了避免图像数据向量化后的维数灾难问题,以及增强对野值(outliers)及噪声的鲁棒性,该文提出一种基于L1-范数的2维线性判别分析(L1-norm-based Two-Dimensional Linear Discriminant Analysis,2DLDA-L1)降维方法。它充分利用L1-范数对野值及噪声的强鲁棒性,并且直接在图像矩阵上进行投影降维。该文还提出一种快速迭代优化算法,并给出了其单调收敛到局部最优的证明。在多个图像数据库上的实验验证了该方法的鲁棒性与高效性。展开更多
块主成份分析(block principal component analysis,BPCA)是一种重要的子空间学习方法,能充分利用图像矩阵的部分关联.基于L1-范数的BPCA是近年来发展起来的鲁棒降维的有效方法.本研究提出了一种新的鲁棒稀疏BPCA方法,称之为BPCAL1-S....块主成份分析(block principal component analysis,BPCA)是一种重要的子空间学习方法,能充分利用图像矩阵的部分关联.基于L1-范数的BPCA是近年来发展起来的鲁棒降维的有效方法.本研究提出了一种新的鲁棒稀疏BPCA方法,称之为BPCAL1-S.该方法相对于传统的基于L2-范数的PCA对噪声更加鲁棒.为了建立稀疏模型,优化过程中引入弹性网,联合使用Lasso与Ridge惩罚因子进行约束.提出了一种贪心算法逐个提取特征向量,对迭代过程的收敛性做了理论证明.将BPCAL1-S应用于图像分类与图像重构,实验结果验证了该方法的有效性.展开更多
The title compound [HphenNO2]+NO3- has been prepared and characterized by elemental analysis, electronic absorption spectroscopy, TG/DTA, IR, 1H and 13C NMR spectro- scopy. Single-crystal X-ray structure determination...The title compound [HphenNO2]+NO3- has been prepared and characterized by elemental analysis, electronic absorption spectroscopy, TG/DTA, IR, 1H and 13C NMR spectro- scopy. Single-crystal X-ray structure determination of the title compound was also carried out. It crystallizes in monoclinic, space group Cc with a = 13.861(3), b = 10.142(2), c = 8.7320(17) ? b = 103.70(3)? C12H8N4O5, Mr = 288.22, V = 1192.6(4) 3, Z = 4, Dc = 1.605 g/cm3 , F(000) = 592, (MoK) = 0.129 mm-1, R = 0.0439, wR = 0.1125 and GOF =1.114. In the crystal lattice, the molecules create a network structure through hydrogen bonds. The second order optical non- linearity was performed by quantum chemical method, showing the title compound has higher molecular hyper polarizability value (?= 24.66×10-30 esu).展开更多
文摘为了避免图像数据向量化后的维数灾难问题,以及增强对野值(outliers)及噪声的鲁棒性,该文提出一种基于L1-范数的2维线性判别分析(L1-norm-based Two-Dimensional Linear Discriminant Analysis,2DLDA-L1)降维方法。它充分利用L1-范数对野值及噪声的强鲁棒性,并且直接在图像矩阵上进行投影降维。该文还提出一种快速迭代优化算法,并给出了其单调收敛到局部最优的证明。在多个图像数据库上的实验验证了该方法的鲁棒性与高效性。
文摘块主成份分析(block principal component analysis,BPCA)是一种重要的子空间学习方法,能充分利用图像矩阵的部分关联.基于L1-范数的BPCA是近年来发展起来的鲁棒降维的有效方法.本研究提出了一种新的鲁棒稀疏BPCA方法,称之为BPCAL1-S.该方法相对于传统的基于L2-范数的PCA对噪声更加鲁棒.为了建立稀疏模型,优化过程中引入弹性网,联合使用Lasso与Ridge惩罚因子进行约束.提出了一种贪心算法逐个提取特征向量,对迭代过程的收敛性做了理论证明.将BPCAL1-S应用于图像分类与图像重构,实验结果验证了该方法的有效性.
文摘The title compound [HphenNO2]+NO3- has been prepared and characterized by elemental analysis, electronic absorption spectroscopy, TG/DTA, IR, 1H and 13C NMR spectro- scopy. Single-crystal X-ray structure determination of the title compound was also carried out. It crystallizes in monoclinic, space group Cc with a = 13.861(3), b = 10.142(2), c = 8.7320(17) ? b = 103.70(3)? C12H8N4O5, Mr = 288.22, V = 1192.6(4) 3, Z = 4, Dc = 1.605 g/cm3 , F(000) = 592, (MoK) = 0.129 mm-1, R = 0.0439, wR = 0.1125 and GOF =1.114. In the crystal lattice, the molecules create a network structure through hydrogen bonds. The second order optical non- linearity was performed by quantum chemical method, showing the title compound has higher molecular hyper polarizability value (?= 24.66×10-30 esu).
