论文部分内容阅读
卷积神经网络(Convolutional Neural Networks-简称CNN)是人工神经网络的一种,由于其具有结构简单、训练参数少、适应性强、效率高等特点,在模式识别、图像处理等领域得到了广泛的应用。本文从CNN的发展历史开始,对其网络结构、神经元模型和训练方法进行了简要描述。在此基础上,实现了一个使用Java语言实现的CNN算法,并给出了算法的的程序框架与代码实现,该算法可以在不使用任何类库的情况下对MINIST手写数字数据集进行分析识别,达到了较高的准确率。
Convolutional Neural Networks (CNN) is a kind of artificial neural network. Because of its simple structure, less training parameters, high adaptability and high efficiency, Convolutional Neural Networks (CNN) has been widely used in the field of pattern recognition and image processing application. This article starts with the history of CNN and briefly describes its network structure, neuron model and training method. On this basis, a CNN algorithm implemented in the Java language is implemented, and the program framework and code implementation of the algorithm are given. The algorithm can analyze and identify the MINIST handwritten digital data set without using any class library , Reached a higher accuracy rate.