自适应图嵌入的鲁棒稀疏局部保持投影  

Robust sparse locality preserving projection with adaptive graph embedding

作　　者：占善华[1,2] 武继刚[1,2] 房小兆[1,2] ZHAN Shan-hua;WU Ji-gang;FANG Xiao-zhao(College of Computer,Guangdong University of Technology,Guangzhou 510006,China;Department of Information Management,Guangdong Justice Police Vocational College,Guangzhou 510520,China)

机构地区：[1]广东工业大学计算机学院,广东广州510006 [2]广东司法警官职业学院信息管理系,广东广州510520

出　　处：《计算机工程与设计》2020年第8期2296-2301,共6页Computer Engineering and Design

基　　金：国家自然科学基金项目(61672171);广东自然科学基金项目(2018B030311007);广东省普通高校青年创新人才类基金项目(2018GkQNCX076);广东省省级科技基金项目(2019B020208001)。

摘　　要：针对复杂高维数据的维度约减问题,提出一种鲁棒的无监督维度约简方法。将自适应的图学习和投影学习融入一个联合学习框架,自适应捕获数据的本质局部结构,以此指导模型学习到全局最优的投影;为捕获数据的全局信息,引入一个PCA项,该项的引入能够减少维度约简过程中的信息损失;为选择最重要的特征进行维度约简,引入一个行稀疏约束,增强对于噪声的鲁棒性。在多个数据集上的实验验证了所提方法的有效性。Aiming at the problems of dimensionality reduction of complex high dimensional data,a robust unsupervised dimensionality reduction method termed robust sparse locality preserving projection with adaptive graph embedding was proposed.The adaptive graph learning and projection learning were integrated into a framework,which captured the intrinsic locality structure of data and in turn promoted the method to achieve the global optimal projection.To capture the global information of data,a variant PCA term was introduced,which decreased the information loss during dimensionality reduction.A row-sparsity constraint was imposed on the projection to select the most important features for dimensionality reduction,so as to improve the robustness of the proposed method to noises.Extensive experiments were performed on multiple databases,which sufficiently validated the superiority of the proposed method in comparison with some state-of-the-art-methods.

关 键 词：投影学习 维度约减 稀疏约束 图嵌入 无监督学习 高维数据 

分 类 号：TP181[自动化与计算机技术—控制理论与控制工程]