Int J Performability Eng ›› 2019, Vol. 15 ›› Issue (9): 2515-2521.doi: 10.23940/ijpe.19.09.p25.25152521

Previous Articles     Next Articles

Natural Image Classification based on Multi-Parameter Optimization in Deep Convolutional Neural Network

Lei Wanga,b,c,*, Yanning Zhanga,b, Runping Xia,b, and Lu Lingc   

  1. aSchool of Computer Northwestern Polytechnical University, Xi'an, 710072, China;
    bNational Engineering Laboratory for Integrated Aero-Space-Ground-Ocean Big Data Application Technology, Xi'an, 710072, China;
    cJiangxi Engineering Technology Research Center of Nuclear Geoscience Data Science and System, East China University of Technology, Nanchang, 330013, China
  • Submitted on ; Revised on ; Accepted on
  • Contact: *. Lei Wang is currently a PhD student at Northwestern Polytechnic University. She received her MSc degree in earth exploration and information processing from East China University of Technology. She is an associate professor in the School of Information Engineering at East China University of Technology. Her current research interests include computer vision, 3D data analysis, and deep learning. E-mail address:

Abstract: Traditional machine learning algorithms cannot adequately train the parameters of networks using massive data. A deep convolutional neural network based on multi-parameter optimization by the TensorFlow deep learning framework is designed in this paper. In order to improve the training speed and prevent over-fitting, we improve and optimize the multi-parameters, including the batch value, dropout, and momentum in the network structure. The experiment involves training and testing on the standard natural image data sets in cifar-10 and cifar-100. The experimental results show that the method achieve better classification accuracy in less time compared with other algorithms, such as Conv-KN, ImageNet-2010, SVM, LR, and Boosting.

Key words: deep convolutional neural network, multi-parameter optimization, batch normalization, dropout, momentum