摘要
分析了平方差损失函数(Quadratic Cost Function)和交叉熵损失函数(Cross Entropy Cost Function)对人工神经网络收敛性的影响,解释了使用平方差损失函数造成网络训练时间较慢的主要原因,并通过仿真实验验证了交叉熵损失函数可提升网络收敛速度.
The paper artifical neural network mainly analyses the impact of Quadratic Cost Function and Cross Entropy Cost Function on the astringency of artificial neural network,and explains the main reason of decreasing the rate of training network caused by Quadratic Cost Function.Finally,the experiment proved that Cross Entropy Cost Function can improve the convergence rate of network.
作者
任进军
王宁
REN Jin-jun;WANG Ning(College of Computer Science and Engineering,Northwest Normal University,Lanzhou Gansu 730070;Department of Science Teaching,Gansu University of Chinese Medicine,Dingxi Gansu 743000)
出处
《甘肃高师学报》
2018年第2期61-63,共3页
Journal of Gansu Normal Colleges
基金
甘肃中医药大学定西校区校级项目"多类别分类算法研究及其在医学图像分类中的应用"(2017XJYB08)
关键词
人工神经网络
平方差损失函数
交叉熵损失函数
网络收敛速度
artificial neural network
quadratic cost function
cross entropy cost function
convergence rate of network
