Optimizing Convolutional Neural Network Architecture Using a Self-adaptive Harmony Search Algorithm

  • Yin-Fu HuangEmail author
  • Jung-Sheng Liu
Conference paper
Part of the Advances in Intelligent Systems and Computing book series (AISC, volume 1074)


In recent years, the advance of GPUs led to the development in neural networks and deep learning. However, it is difficult to find a good CNN architecture people desire. In the past, people had to manually find the CNN architecture, and this is quite time-consuming and labor-intensive. In this paper, we use a self-adaptive harmony search algorithm to find the optimized convolutional neural network architecture for image recognition. The system architecture is divided into two phases. In the first phase, we search for the most suitable layer length of a CNN. In the second phase, we fine-tune the architecture found in the first phase or a pre-trained architecture. In the experiments, three popular and well-known datasets are used to evaluate the proposed methods and the state-of-the-art CNN search methods. The experimental results show that our methods achieve competitive performances compared with the other methods on CIFAR-10, and have the best performance among all the methods on MNIST and Caltech-101.


Convolutional Neural Networks Deep learning Hyper-parameter optimization Self-adaptive harmony search algorithms Transfer learning 


  1. 1.
    Albelwi, S., Mahmood, A.: Automated optimal architecture of deep convolutional neural networks for image recognition. In: Proceedings of the 15th IEEE International Conference on Machine Learning and Applications, pp. 53–60 (2016)Google Scholar
  2. 2.
    Al-Hyari, A., Areibi, S.: Design space exploration of convolutional neural networks based on evolutionary algorithms. J. Comput. Vis. Imaging Syst. 3(1) (2017)Google Scholar
  3. 3.
    Bottou, L.: Large-scale machine learning with stochastic gradient descent. In: Proceedings of the 19th International Conference on Computational Statistics, pp. 177–186 (2010)Google Scholar
  4. 4.
    Ioffe, S., Szegedy, C.: Batch normalization: accelerating deep network training by reducing internal covariate shift. arXiv:1502.03167v3, pp. 1–11 (2015)
  5. 5.
    Khan, H.A.: DM-L based feature extraction and classifier ensemble for object recognition. J. Sig. Inf. Process. 9(2), 92–110 (2018)Google Scholar
  6. 6.
    Krogh, A., Hertz, J.A.: A simple weight decay can improve generalization. Adv. Neural. Inf. Process. Syst. 4, 950–957 (1992)Google Scholar
  7. 7.
    Lin, M., Chen, Q., Yan, S.: Network in network. arXiv:1312.1400v3, pp. 1–10 (2014)
  8. 8.
    Pham, H., Guan, M.Y., Zoph, B., Le, Q.V., Dean, J.: Efficient neural architecture search via parameter sharing. arXiv:1802.03268v2 (2018)
  9. 9.
    Rosa, G., Papa, J., Marana, A., Scheirer, W., Cox, D.: Fine-tuning convolutional neural networks using harmony search. In: Proceedings the 20th Iberoamerican Congress on Pattern Recognition, pp. 683–690 (2015)Google Scholar
  10. 10.
    Simonyan, K., Zisserman, A.: Very deep convolutional networks for large-scale image recognition. arXiv:1409.1556v6, pp. 1–14 (2015)
  11. 11.
    Singh, A., Kingsbury, N.: Efficient convolutional network learning using parametric log based dual-tree wavelet scatternet. In: Proceedings of the IEEE International Conference on Computer Vision Workshops, pp. 1140–1147 (2017)Google Scholar
  12. 12.
    Szegedy, C., Vanhoucke, V., Ioffe, S., Shlens, J., Wojna, Z.: Rethinking the inception architecture for computer vision. In: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pp. 2818–2826 (2016)Google Scholar
  13. 13.
    Tanno, R., Arulkumaran, K., Alexander, D.C., Criminisi, A., Nori, A.: Adaptive neural trees. arXiv:1807.06699v2 (2018)
  14. 14.
    Wang, C.M., Huang, Y.F.: Self-adaptive harmony search algorithm for optimization. Expert Syst. Appl. 37(4), 2826–2837 (2010)CrossRefGoogle Scholar
  15. 15.
    Young, S.R., Rose, D.C., Karnowski, T.P., Lim, S.H., Patton, R.M.: Optimizing deep learning hyper-parameters through an evolutionary algorithm. In: Proceedings of the Workshop on Machine Learning in High-Performance Computing Environments, Article No. 4 (2015)Google Scholar
  16. 16.
    Zeiler, M.D.: ADADELTA: an adaptive learning rate method. arXiv:1212.5701v1 (2012)

Copyright information

© Springer Nature Switzerland AG 2020

Authors and Affiliations

  1. 1.National Yunlin University of Science and TechnologyTouliuTaiwan, R.O.C.

Personalised recommendations