一种改进深度学习网络结构的英文字符识别

An Improved Deep Learning Network Structure for English Character Recognition

自Geoffrey Hinton于2006年在《Reducing the dimensionality of data with neural networks》一文中首次提出深度学习(Deep Learning)的概念,深度学习就受到了研究人员的持续关注。深度学习利用多层的神经网络模拟人类大脑的多层抽象学习过程。其中网络结构设计和特征提取是数据挖掘和模式识别应用中的关键问题。而深度学习对手写体数字识别的准确率一直是衡量一个深度学习算法或网络结构优劣的重要标准。提出一种改进的深度学习网络结构,通过对手写体英文数据库Letter Recognition的识别实验结果表明,该深度学习网络结构的识别正确率相比传统的深度学习网络有了明显的提高。

Since Geoffrey Hinton proposed the concept of ＂ Deep Learning＂ in a paper, ＂ Reducing the dimensionality of data with neural networks＂,in 2006,in the first time. Deep learning has received sustained attention from researchers.Deep learning use multi-layer neural network to simulation of the multi-layer abstract learning process of human brains.The design of network structure and feature extraction are the key problems in data mining and applications of pattern recognition. The accuracy of handwritten numeral recognition in deep learning has always been an important criterion for measuring the deep learning algorithm or network structure. In this paper,an improved network of deep learning structure is proposed. The experimental results show that the recognition rate of the deep learning network structure is significantly higher than the traditional network structure deep learning.