摘要
跨层卷积神经网络模型由输入层、两个交替的卷积层、池化层、全连接层和输出层组成。池化层输出到全连接层,将网络的高层次特征和低层次特征相结合构造分类器。在网络中加入Dropout技术,以防止过拟合的发生。
A cross-layer convolution neural network model is composed of input,two staggered convolution,pooling,full connection and output layers.The output of the two pooling layers are sent to the full connection layer for building the classifier with both high-level features and low-level features of the network.Dropout technology is added to the network for preventing over-fitting.
作者
崔明光
张秀梅
韩维娜
CUI Mingguang;ZHANG Xiumei;HAN Weina(School of Electrical&Electronic Engineering,Changchun University,Changchun 130012,China)
出处
《长春工业大学学报》
CAS
2019年第4期332-338,共7页
Journal of Changchun University of Technology
基金
国家自然科学基金资助项目(61374051)
关键词
卷积神经网络
缺陷识别
跨层连接
Dropout技术
convolutional neural network
defect recognition
cross-layer connection
Dropout technology