基于多目标优化问题模型的鉴别分析方法

该文提出一种新的Fisher鉴别分析的多目标优化问题模型。模型将最大的Fisher函数值与鉴别矢量所抽取特征间最小的相关同时作为优化目标。通过构造评价函数，将这一问题转换为适当的单目标优化问题。利用问题模型的特殊性，用Lagrange乘子法分析得出了相应鉴别矢量解的表达式。将该方法与F—S鉴别分析方法在CENPARMI手写体阿拉伯数字库上进行了对比实验。实验显示，该文方法明显优于F—S鉴别分析方法。

Discriminant Analysis Method Based on Multi-goal Optimization Model

A multi-goal optimization model for Fisher discriminant analysis is developed. In the new model, maximal Fisher criterion values of discriminant vectors and minimal statistical correlation between the features extracted by the discriminant vectors are simultaneously required. Then, the original problem model is transformed into a single-goal model. The discriminant vectors corresponding to the latter model are worked out by Lagrange method. An experiment on CENPARMI handwritten numeral database is performed. By contrast with F-S discriminant analysis, the experiment shows that the method is much more effective.