摘要
文章介绍了一种利用BP神经网络进行数字识别的方法。首先,对数字进行特征提取,获得采样数据,再对样本数据进行学习和训练,形成良好的网络,然后对与训练数字有所区别的数字进行检测,达到了一定的准确度,表明了该方法在实际应用中具有可行性。本文共分为五部分,第一部分对神经网络的基本原理进行了简单介绍,第二部分讲述了反向传播算法的基本原理,第三部分讲述了数字识别的基本原理,第四部分讲述了基于人工神经网络的数字识别的实例,第五部分对上述内容作了简要小结。
This paper introduces an approach to identify the digits based on BP neural networks. Firstly, the data of specimen is obtained, and then learning the data to form a perfect structure of a neural network. Secondly, the perfect neural network is used to detects digits, which are different from the trained digits. The detecting results of the networks can reach the required accuracy of digital identification, which show the practical application feasibility of this approach is.
出处
《物探化探计算技术》
CAS
CSCD
2005年第1期78-80,99-100,共3页
Computing Techniques For Geophysical and Geochemical Exploration
基金
国家"十五"科技攻关项目(2001BA601B05)
