中图法分类号TP391.doc

中图法分类号: TP391.4 文献标识码:A 非监督子空间学习中关联度量 多维尺度分析1）本文得到国家自然科学基金（1010200720100098）资助； 2）收稿日期：2010-9-15 3）何光辉(1976.8- )，男，讲师，2010年12月获得重庆大学计算机应用与技术专业博士学位。主要研究方向模式识别、数据挖掘，E-mail：ghhe@ 何光辉1) ，张太平2) ，唐远炎2) 1)重庆大学数学与统计学院 401331，重庆 2)重庆大学计算机学院 400044，重庆 摘要：多维尺度分析已经在维度约减和数据挖掘领域得到了广泛应用。MDS的主要缺点是其定义在训练数据上，对于新的测试样本无法直接获得映射结果。另外，MDS基于欧氏距离度量，不适合获取相似数据中的非线性流形结构。本文将MDS扩展到关联度量空间，称为关联度量多维尺度分析（Correlation MDS，CMDS）。与传统MDS在训练数据中完成映射，进而缩小空间范围相比，CMDS 能够直接获得测试样本映射结果。此外，CMDS基于关联度量，能够有效学习相似数据中的非线性流形结构。理论分析表明，CMDS可以利用核方法扩展到新特征空间，解决非线性问题。实验证明CMDS及其核形式KG-CMDS性能优于常用传统降维方法。 关键字：非监督子空间；关联度量；多维尺度分析 Correlation-based multidimensional scaling for unsupervised subspace learning Guanghui He1)，Taiping Zhang2)，Yuanyan Tang2) 1)College of mathematics and statistics ,Chongqing University,401331，Chongqing ; 2)College computer science ,Chongqing University,400044，Chongqing Abstract：Multidimensional scaling (MDS) has been applied in many applications such as dimensionality reduction and data mining. However, one of the drawbacks of MDS is that it is only defined on “training” data with no clear extension to out-of-sample points. Furthermore, since MDS is based on Euclidean distance, it is not suitable for detecting the nonlinear manifold structure embedded in the similarities between data points. In this paper, we extend MDS to the correlation measure space(CMDS). In contrast with MDS where the mapping between the input and reduced spaces is implicit, CMDS employs an explicit nonlinear mapping between the both. As a result, CMDS can directly provide prediction for new samples. In addition, correlation is a similarity measure, CMDS method can effectively capture the nonlinear manifold structure of data embedded in the similarities between the data points. Theoretical analysis also shows that CMDS has some properties similar to kernel methods and can be extended to feature space. The effectiveness of our approach is demonstrated by extensive experiments on various data sets, in comparison with existing dimensi