手写数字特征提取与选择的研究
Handwritten digital feature extraction and selection
本文提出5种共60维手写数字的典型特征集，并使用k-w分类准则函数和直方图方法对特征集进行选择，最终选择其中的33维，达到了降低输入维数的目的。实验表明，提取的特征具有很好的分类效果，并且，降低特征输入的维数显著地提升了神经网络的学习效率。
his paper probes five kinds, 60 dimension typical features set of digits, and uses the k-w classification criteria function and histogram methods to select 33 features of them, which reduce the input dimension. Experiments show that the selected features have better classification performance, and dimensionality reduction of feature input greatly improved the efficiency of neural network learning.
王昱、康明
计算技术、计算机技术
特征提取特征选择BP网络
feature extractionfeature selectionBP networks
王昱,康明.手写数字特征提取与选择的研究[EB/OL].(2010-04-26)[2025-08-02].http://www.paper.edu.cn/releasepaper/content/201004-930.点此复制
评论