论文部分内容阅读
本文中的二分法是指:把多分类问题逐级化为二分类问题;用结构简单的反传网络群代替结构复杂的单一网络;在用大样本对反传网络进行分类训练时分两步进行,第一步粗精度,使训练很快完成,然后根据网络的分类结果重新确定样本的类别,这样之后,第二步很快就可使训练达到高精度。
The dichotomy in this paper refers to the classification of multi-classification problems into two classification problems; the simple structure of the anti-transmission network instead of the complex structure of a single network; large samples of the anti-transmission network classification training in two steps , The first step of coarse precision, so that the training is completed quickly, and then re-determine the type of samples according to the classification results of the network, so that the second step quickly to make training to achieve high accuracy.