保留复杂化学模式特征的映射及其应用  

The feature-preserving map of complex chemical patterns and its application

作　　者：颜学峰[1] 余娟[1] 钱锋[1] 丁军委[2] 陈德钊[2] 胡上序[2] 

机构地区：[1]华东理工大学自动化研究所,上海200237 [2]浙江大学化工系计算机仿真教研室,浙江杭州310027

出　　处：《计算机与应用化学》2005年第11期1013-1017,共5页Computers and Applied Chemistry

基　　金：国家自然科学基金(20506003)上海启明星项目(04QMX1433)国家973计划(2002CB312200)国家863计划(2002AA412110)

摘　　要：提出一种保留模式特征的映射算法,并具体应用于高维的、分量间存在复共线性的复杂化学模式特征的映射,取得良好效果。算法首先提取能最大限度表达复杂化学模式特征的特征变量,并进而采用自组织映射算法将由特征变量组成的特征模式映射到低维空间,使自组织映射结果平面能直观地显示复杂化学模式的特征。算法在具体应用过程中根据复杂化学模式样本的特性,提出了主成分特征映射和分类相关成分特征映射算法。A novel map approach, named as pattern feature-preserving map, was proposed to obtain the two-dimensional feature-preserving map of the high-dimensional patterns and to avoid from the troubles that some independent variables have only a little or not effect on the pattern feature and that there exists significant correlation among some independent variables. The two-dimensional feature-preserving map can more concisely and efficiently represent the pattern feature. Firstly, the feature variables, which are derived from the original independent variables, have no correlation with each other, and can represent the pattern feature at the utmost. Then, the feature variables are employed as the input variables for the self-organizing map networks. In the practical applications, principal component analysis and correlative component analysis were employed to identify the feature variables for the feature-preserving map of the complex patterns.

关 键 词：复杂化学模式 特征映射 主成分 分类相关成分 自组织映射 

分 类 号：O212.4[理学—概率论与数理统计] O235[理学—数学]