欢迎您！东篱公司
退出
-
申报数据库
-
立项数据库
-
成果数据库
-
项目获奖数据库
中文摘要：
为了在保持数据局部几何结构不变的同时使数据分类边界最大化，提出了一种用于分类的线性局部切空间判别分析算法．该算法是改进的流形学习算法的监督版，样本的局部切空间排列矩阵确保样本低维嵌入的局部几何结构不变；基于最大边界准则的数据散度矩阵确保数据分类的类内散度最小和类间散度最大；对上述2个矩阵和进行特征分解，获得平衡的投影向量基，使样本投影后的子空间被优化．对Yale，UMIST与MIT这3个人脸数据库的实验结果表明，与现有多种经典分类方法相比，提出的算法在降维的同时提取了用于人脸识别的更有效特征，识别性能较好，具有较高的判别分析能力．
英文摘要：
A novel algorithm, called linear local tangent space discriminant analysis （LLTSDA）, is proposed for classification, which is motivated by the desire to preserve the local geometry structure of data and to maximize the classification margin of data. The proposed algorithm is a supervised and modified manifold learning algorithm. The matrix of local tangent space alignment preserves the local geometry structure of data in low-dimensional embedded space; while the scatter matrix can ensure the within-class scatter minimization and the between-class scatter maximization; and a trade-off projected vector base can be taken by solving the generalized eigenvalue problem to the summation of above both matrices. Compared with classical classification methods, the experimental results from Yale, UMIST （University of Manchester Institute of Science and Technology） and MIT face databas- es show that LLTSDA algorithm can extract the more efficient features for face recognition while the dimensionality is reduced, and obtains much higher recognition accuracies and stronger classification power,
同期刊论文项目
同项目期刊论文
EpiMiner: a three-stage co-information based method for detecting and visualizing epistatic interact
Statistical tests based multi-strategy learning method to infer the Gaussian Markov network structur
Probability Theory-Based SNP Association Study Method for Identifying Susceptibility Loci and Geneti
An overall-regional competitive self-organizing map neural network for the Euclidean traveling sales
A Computational Method Based on the Integration of Heterogeneous Networks for Predicting Disease-Gen
Seed selection strategy in global network alignment without destroying the entire structures of func
Predicting protein complexes in protein interaction networks using a core-attachment algorithm based
A degree-distribution based hierarchical agglomerative clustering algorithm for protein complexes id
Identification of overlapping and non-overlapping community structure by fuzzy clustering in complex
Scaling the Kernel Function based on the Separating Region in Input Space: A Data-dependent Way to I
Non-traditional spectral clustering algorithms for the detection of community structure in complex n
A network based method for analysis of lncRNA-disease associations and prediction of lncRNAs implica
期刊信息
位置：