摘要
贝叶斯分类器、线性分类器和K近邻分类器是模式识别中三种典型的模式分类器。比较三种分类器在识别手写阿拉伯数字过程中的性能优缺点,进一步对识别数据进行详尽的分析挖掘,通过对算法精确度、识别速度及计算存储需求等方面的比较,深入探讨三种监督式分类器的差异和特点,最终得到不同的分类结果,从而寻求最优化的决策方案。
Bayes classifier, linear classifier and K-nearest neighbor classifier are the typical supervised learning classifiers in pattern classification. In this paper, the advantages and disadvantages of these classification rules are discussed in terms of handwritten digit recognition problem. We explore the handwritten digit database and compare the performance of different classifiers in terms of error rate and computation cost. The results show that the error rate is lower for the bays rule and the K-NN than that of Linear classifier; however, the linear classifier is easier to implement and widely used.
出处
《装备制造技术》
2008年第8期111-114,共4页
Equipment Manufacturing Technology
关键词
手写数字识别
贝叶斯决策
K近邻决策
线性分类器
误差率
handwritten digit recognition, Bayes rule, K-nearest neighbor rule, linear classifier, error rate
