智能分类器方法

Intelligence Classifier Method

小样本数据不能对分类器进行充分的训练。传统的分类器方法一旦设计好,再也不会有任何改进.本文将人工智能的思想、方法应用于分类器设计中,提出了智能分类器概念。智能分类器不但可以对未知样本进行分类识别,还具有多专家决策、预分类、学习等功能。智能分类器是一种自适应系统,其系统参数可在识别的过程中得到不断的优化。在ORL人脸库上的实验结果证明该方法在解决小样本问题时具有明显的优势。

The sample database with small size can not be used to train the classifiers sufficiently and the traditional classifiers may not have any change once they are designed. The artificial intelligence idea and methods are considered to design the classifiers and the concept of the intelligence classifier（ICF） is proposed. ICF can classify unknown samples as the traditional classifier. It also has some functions such as multi-experts decision, pre-classifying and learning. So ICF is a self-adaptive system whose parameters can be optimized in classification. The experimental results on ORL face database show that the intelligent classifier has great advantages in solving the problem of small sample size.