基于流形保持的半监督局部排列降维算法  

De Semi-supervised Dimensionality Reduction Based on Manifold Structure Preserving and Discriminative Locality Aligment

作　　者：宋德华[1] 姚莉秀[1] 

机构地区：[1]上海交通大学电子信息与电气工程学院

出　　处：《微型电脑应用》2013年第5期17-20,共4页Microcomputer Applications

摘　　要：判别局部排列是基于谱分析片排列框架下的降维算法,但是,算法只能针对单流形数据进行降维。针对判别局部排列算法存在的缺陷,着重研究了多流形学习和半监督学习技术,利用标签传播算法(LP)和线性重构分析,提出一种流行结构保持的半监督降维算法,利用标签传播后得到的全体样本标签信息进行片都构建,并通过求解目标函数的最优解来获得低维嵌入。在YALE和FERET这两个标准人连数据库上的实验,验证了算法的有效性能并体现了算法在分类上的良好性能。While bearing the benefit of time efficiency, IMRT techniques face the deterioration of plan quality due to minimization of intensity modulation in exchange for fast delivery of treatment plan. In order to achieve an optimal compromise between plan quality and delivery efficiency, a variant of adaptive IMRT（A-IMRT） treatment planning technique is proposed in this study, which provides a solution to calculate a single-arc IMRT plan with the minimized numbers of beams of beams and the minimized levels of their intensities. Our results show that the proposed A-IMRT optimization technique provides an effective way to reach a single-arc plan with the best compromise between dose quality and delivery efficiency and competent for the applications of the upcoming surge of rapid delivery of treatment plan on the simulated abdomen phantom.

关 键 词：DLA 标签传播算法 多流行学习 半监督学习 人脸识别 

分 类 号：TP311[自动化与计算机技术—计算机软件与理论]