补缀式浓缩近邻分类器BDPATCH
A CNN Classification Design with Boundary Patching
-
摘要: 本文提出一种浓缩近邻分类器BDPATCH.其浓缩集从编辑过的训练集经显露和补缀边 界模式产生,具有Bayes渐近最优性.对BDPATCH和其它已有的CNN算法进行了比较, 结果表明这种新的分类器具有高识别率,同时又是快速的.Abstract: A condensed nearest neighbor classification rule, BDPATCH, is proposed. The condensed set is produced by exposing and patching the boundary pattern subset of the edited training set. This procedure results in a Bayes asymptotically optimal classifier. Simulation experiment on this new and other existing CNN rules is presented, which shows that BDPATCH can achieve a high recognition rate and maintain a fast speed.
计量
- 文章访问数: 1299
- HTML全文浏览量: 44
- PDF下载量: 1151
- 被引次数: 0