排序方式: 共有3条查询结果,搜索用时 0 毫秒
1
1.
针对现有稀疏子空间聚类算法获取的系数矩阵不能准确反应高维空间中数据分布的稀疏性的不足,提出一种分式函数约束的稀疏子空间聚类模型,并利用交替方向迭代方法给出该模型的解。在无噪声情形下,证明了该方法获取的系数矩阵具有块对角结构,这为其准确获取数据结构提供了理论保证;在含噪声情形下,对异常点噪声同样采用分式函数约束作为正则项,提高了模型的鲁棒性。在人工数据集、Extended Yale B库和Hopkins155数据集上的实验结果表明,基于分式函数约束的稀疏子空间聚类方法不仅提高了聚类结果的准确率,而且对异常点噪声具有更好的鲁棒性。 相似文献
2.
In recent years, magnetic interference thin films have gained wide attention in optical security devices field by virtue of their gonioapparent and dynamic 3D effects. Based on the color mechanism of metal-dielectric Fabry-Perot structure, a novel seven-layer magnetic thin film structure is proposed by adopting the ultrathin metal layer as a bonding layer and a pure metallic Ni layer as a magnetic layer as well as a reflective layer. Color target optimization optimac method is utilized that realizes the seven-layer metal-dielectric optically variable magnetic thin film structure with green at the normal incidence and purple-red at 60°. The structure effectively solves the delaminate problem and simplifies the multilayer structure. Through different combined magnetic field designs, the magnetic orientation experiment of the prepared magnetic optically variable thin film is carried out, and the 3 D anti-counterfeiting with remarkable dynamic color change effect is obtained, which provides a new solution for optical security devices. 相似文献
3.
1