论文部分内容阅读
本文在论述模式识别的统计方法和模糊方法的共同性、差异以及各自适用范围的基础上,研究了模式识别的统计模糊方法和模糊统计方法。统计模糊方法是在模糊分类器中充分利用模式分量统计信息的隶属函数,使分类性能优于普通的模糊分类器。模糊统计方法是在以统计方法为基础的分类器中,用模式分量的模糊隶属函数代替模式分量作为分类器输入。从对本文中几个数据集所作的分类试验结果看,这种方法只需要不大的训练样本集便可使分类性能接近于Bayes分类器的最佳水平。
Based on the discussion of the commonality and difference of statistical methods and fuzzy methods of pattern recognition and the respective applicable scopes, this paper studies the statistical fuzzy method and fuzzy statistical method of pattern recognition. The statistical fuzzy method makes full use of the membership function of the statistical information of the mode component in the fuzzy classifier so that the classification performance is better than that of the ordinary fuzzy classifier. Fuzzy statistical method is based on the statistical method based on the classifier, fuzzy membership function of mode components instead of mode components as a classifier input. According to the classification test results of several data sets in this paper, this method requires only a small training sample set to bring the classification performance close to the best level of Bayesian classifier.