Enhancement network: confining deep convolutional features based on hand-crafted rules

doi:10.19682/j.cnki.1005-8885.2020.0035

中国邮电高校学报(英文) ›› 2020, Vol. 27 ›› Issue (4): 34-42.doi: 10.19682/j.cnki.1005-8885.2020.0035

• Artificial Intelligence • 上一篇下一篇

Enhancement network: confining deep convolutional features based on hand-crafted rules

李艳华肖文光

1. 安徽工商职业学院
2. 中国电子科技集团公司第三十八研究所

收稿日期:2020-06-05 修回日期:2020-08-09 出版日期:2020-08-31 发布日期:2020-08-31
通讯作者: 肖文光 E-mail:xwgb711@163.com
基金资助:
安徽省高等学校自然科学研究重点项目

Enhancement network: confining deep convolutional features based on hand-crafted rules

Li Yanhua, Xiao Wenguang

Received:2020-06-05 Revised:2020-08-09 Online:2020-08-31 Published:2020-08-31
Supported by:
Key Project of Natural Science Research in Anhui Province

摘要/Abstract

摘要：

Despite convolutional neural network ( CNN) is mature in many domains, the understanding of the directions where the parameters of the CNNs are learned towards, falls behind, and researches on the functions that the convolutional networks (ConvNets) learns are difficult to be explored. A method is proposed to guide ConvNets to learn towards the expected direction. First, for the sake of facilitating network converging, a novel feature enhancement framework, namely enhancement network (EN), is devised to learn parameters according to certain rules. Second, two types of hand-crafted rules, namely feature-sharpening (FS) and feature-amplifying (FA) are proposed to enable effective ENs, meanwhile are embedded into the CNN for the end-to-end learning. Specifically, the former is a tool sharpening convolutional features and the latter is the one amplifying convolutional features linearly. Both tools aim at the same spot achieving a stronger inductive bias and more straightforward loss functions. Finally, the experiments are conducted on the mixed National Institute of Standards and Technology (MNIST) and cooperative institute for Alaska research 10 (CIFAR10) dataset. Experimental results demonstrate that ENs make a faster convergence by formulating hand-crafted rules.

关键词:

convolutional neural network, network convergence, hand-crafted rules

Abstract: Despite convolutional neural network ( CNN) is mature in many domains, the understanding of the directions where the parameters of the CNNs are learned towards, falls behind, and researches on the functions that the convolutional networks (ConvNets) learns are difficult to be explored. A method is proposed to guide ConvNets to learn towards the expected direction. First, for the sake of facilitating network converging, a novel feature enhancement framework, namely enhancement network (EN), is devised to learn parameters according to certain rules. Second, two types of hand-crafted rules, namely feature-sharpening (FS) and feature-amplifying (FA) are proposed to enable effective ENs, meanwhile are embedded into the CNN for the end-to-end learning. Specifically, the former is a tool sharpening convolutional features and the latter is the one amplifying convolutional features linearly. Both tools aim at the same spot achieving a stronger inductive bias and more straightforward loss functions. Finally, the experiments are conducted on the mixed National Institute of Standards and Technology (MNIST) and cooperative institute for Alaska research 10 (CIFAR10) dataset. Experimental results demonstrate that ENs make a faster convergence by formulating hand-crafted rules.

Key words: convolutional neural network, network convergence, hand-crafted rules

中图分类号:

TN711.3

Li Yanhua, Xiao Wenguang. Enhancement network: confining deep convolutional features based on hand-crafted rules[J]. The Journal of China Universities of Posts and Telecommunications, 2020, 27(4): 34-42.

参考文献

1. Zeiler M D, Fergus R. Visualizing and understanding convolutional networks. Computer Vision: Proceedings of the 14th European Conference on Computer Vision (ECCV'14), 2014, Sept 6-12, Zurich, Switzerland. LNCS 8689. Berlin, Germany: Springer, 2014: 818-833.

2. Bilen H, Fernando B, Gavves E, et al. Dynamic image networks for action recognition. Proceedings of the 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR'16), 2016, Jun 27-30, Las Vegas, NV, USA. Piscataway, NJ, USA: IEEE, 2016: 3034-3042.

3. Berg A, Deng J, Li F F. Large scale visual recognition challenge 2010. ImageNet. Palo Alto, CA, USA: Stanford Vision Lab, Stanford University, 2010

4. Russakovsky O, Deng J, Su H, et al. ImageNet large scale visual recognition challenge. International Journal of Computer Vision, 2015, 115(3): 211-252.

5. Krizhevsky A, Sutskever I, Hinton G E. ImageNet classification with deep convolutional neural networks. Proceedings of the 25th International Conference on Neural Information Processing Systems (NIPS'12): Vol 1, 2012, Dec 3-6, Lake Tahoe, CA, USA. Red Hook, NY, USA: Curran Associates, 2012: 1097-1105.

6. Wu M Y, Li C. Image recognition based on deep learning. Proceedings of the 2015 Chinese Automation Congress (CAC’15), 2015, Nov 27-29, Wuhan, China. Piscataway, NJ, USA: IEEE, 2015: 542-546.

7. Szegedy C, Liu W, Jia Y P, et al. Going deeper with convolutions. Proceedings of the 2015 IEEE Conference on Computer Vision and Pattern Recognition (CVPR’15), 2015, Jun 7-12, Boston, MA, USA. Piscataway, NJ, USA: IEEE, 2015: 1-9.

8. Simonyan K, Zisserman A. Very deep convolutional networks for large-scale image recognition. arXiv 1409.1556, 2014,

9. Sermanet P, Eigen D, Zhang X, et al. OverFeat: Integrated recognition, localization and detection using convolutional networks. arXiv:1312.6229, 2013.

10. Girshick R, Donahue J, Darrell T, et al. Rich feature hierarchies for accurate object detection and semantic segmentation. Proceedings of the 2014 IEEE Conference on Computer Vision and Pattern Recognition, 2014, Jun 23-28, Columbus, OH, USA. Piscataway, NJ, USA: IEEE, 2014: 580-587.

11. Farabet C, Couprie C, Najman L, et al. Learning hierarchical features for scene labeling. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2013, 35(8): 1915-1929.

12. Karpathy A, Toderici G, Shetty S, et al. Large-scale video classification with convolutional neural networks. Proceedings of the 2014 IEEE Conference on Computer Vision and Pattern Recognition, 2014, Jun 23-28, Columbus, OH, USA. Piscataway, NJ, USA: IEEE, 2014: 1725-1732.

13. Zhu Z G, Ji H B, Zhang W B. Nonlinear gated channels networks for action recognition. Neurocomputing, 2020, 386: 325-332

14. Annane D, Chevrolet J C, Chevret S, et al. Two-stream convolutional networks for action recognition in videos. Proceedings of the 27th International Conference on Neural Information Processing Systems (NIPS'14): Vol 1, 2014, Dec 8-13, Montreal, Canada. Cambridge, MA, USA: MIT Press, 2014: 568-576.

15. Hinton G E, Srivastava N, Krizhevsky A, et al. Improving neural networks by preventing co-adaptation of feature detectors. arXiv:1207.0580, 2012.

16. Glorot X, Bordes A, Bengio Y. Deep sparse rectifier neural networks. Proceedings of the 14th International Conference on Artificial Intelligence and Statisitics (AISTATS’11), 2011, Apr 11-13, Ft Lauderdale, FL, USA. 2011: 315-323.

17. Ioffe S, Szegedy C. Batch normalization: Accelerating deep network training by reducing internal covariate shift. Proceedings of the 32nd International Conference on Machine Learning (ICML’15), 2015, Jul 6-11, Lille, France.2015: 448-456.

18. Wang L M, Qiao Y, Tang X O. Action recognition with trajectory-pooled deep-convolutional descriptors. Proceedings of the 2015 IEEE Conference on Computer Vision and Pattern Recognition (CVPR’15), 2015, Jun 7-12, Boston, MA, USA. Piscataway, NJ, USA: IEEE, 2015: 4305-4314.

19. Sun C, Nevatia R. Large-scale Web video event classification by use of Fisher vectors. Proceedings of the 2013 IEEE Workshop on Applications of Computer Vision (WACV’13), 2013, Jan 15-17, Tampa, FL, USA. Piscataway, NJ, USA: IEEE, 2013: 15-22.

20. Zhou J L, Li Z L, Zhi W M, et al. Using convolutional neural networks and transfer learning for bone age classification. Proceedings of the 2017 International Conference on Digital Image Computing: Techniques and Applications (DICTA’17), 2017, Nov 29-Dec 1, Sydney, Australia. Piscataway, NJ, USA: IEEE, 2017: 17-22.

21. Mohammadnia-Qaraei M R, Monsefi R, Ghiasi-Shirazi K. Convolutional kernel networks based on a convex combination of cosine kernels. Pattern Recognition Letters, 2018, 116: 127-134.

22. Krizhevsky A, Hinton G. Learning multiple layers of features from tiny images. Technical Report. Toronto, Canada: University of Toronto, 2009.

23. Jia Y P, Shelhamer E, Donahue J, et al. Caffe: Convolutional architecture for fast feature embedding. Proceedings of the 22nd ACM International Conference on Multimedia (MM'14), 2014, Nov 3-7, Orlando, FL, USA. New York, NY, USA: ACM, 2014: 675-678.

Enhancement network: confining deep convolutional features based on hand-crafted rules

Enhancement network: confining deep convolutional features based on hand-crafted rules

PDF

可视化

摘要/Abstract

引用本文

使用本文

参考文献

相关文章 0

编辑推荐

Metrics

本文评价