半监督深度学习图像分类方法研究综述

doi:10.3778/j.issn.1673-9418.2011020

摘要/Abstract

摘要：

作为人工智能领域近十年来最受关注的技术之一，深度学习在诸多应用中取得了优异的效果，但目前的学习策略严重依赖大量的有标记数据。在许多实际问题中，获得众多有标记的训练数据并不可行，因此加大了模型的训练难度，但容易获得大量无标记的数据。半监督学习充分利用无标记数据，提供了在有限标记数据条件下提高模型性能的解决思路和有效方法，在图像分类任务中达到了很高的识别精准度。首先对于半监督学习进行概述，然后介绍了分类算法中常用的基本思想，重点对近年来基于半监督深度学习框架的图像分类方法，包括多视图训练、一致性正则、多样混合和半监督生成对抗网络进行全面的综述，总结多种方法共有的技术，分析比较不同方法的实验效果差异，最后思考当前存在的问题并展望未来可行的研究方向。

关键词: 半监督深度学习, 多视图训练, 一致性正则, 多样混合, 半监督生成对抗网络

Abstract:

As one of the most concerned technologies in the field of artificial intelligence in recent ten years, deep learning has achieved excellent results in many applications, but the current learning strategies rely heavily on a large number of labeled data. In many practical problems, it is not feasible to obtain a large number of labeled training data, so it increases the training difficulty of the model. But it is easy to obtain a large number of unlabeled data. Semi-supervised learning makes full use of unlabeled data, provides solutions and effective methods to improve the performance of the model under the condition of limited labeled data, and achieves high recognition accuracy in the task of image classification. This paper first gives an overview of semi-supervised learning, and then introduces the basic ideas commonly used in classification algorithms. It focuses on the comprehensive review of image classification methods based on semi-supervised deep learning framework in recent years, including multi- view training, consistency regularization, diversity mixing and semi-supervised generative adversarial networks. It summarizes the common technologies of various methods, analyzes and compares the differences of experimental results of different methods. Finally, this paper thinks about the existing problems and looks forward to the feasible research direction in the future.

Key words: semi-supervised deep learning, multi-view training, consistency regularization, diversity mixing, semi-supervised generative adversarial networks

吕昊远, 俞璐, 周星宇, 邓祥. 半监督深度学习图像分类方法研究综述[J]. 计算机科学与探索, 2021, 15(6): 1038-1048.

LYU Haoyuan, YU Lu, ZHOU Xingyu, DENG Xiang. Review of Semi-supervised Deep Learning Image Classification Methods[J]. Journal of Frontiers of Computer Science and Technology, 2021, 15(6): 1038-1048.

参考文献

[1] ZHANG S, GONG Y H, WANG J J. The development of deep convolution neural network and its applications on com-puter vision[J]. Journal of Computers, 2019, 42(3): 453-482.
张顺, 龚怡宏, 王进军. 深度卷积神经网络的发展及其在计算机视觉领域的应用[J]. 计算机学报, 2019, 42(3): 453-482.
[2] ZHOU F Y, JIN L P, DONG J. Review of convolutional neural network[J]. Journal of Computers, 2017, 40(6): 1229-1251.
周飞燕, 金林鹏, 董军. 卷积神经网络研究综述[J]. 计算机学报, 2017, 40(6): 1229-1251.
[3] CHAPELLE O, SCHOLKOPF B, ZIEN A. Semi-supervised learning[M]. Cambridge: MIT Press, 2006: 34-36.
[4] MEY A, LOOG M. Improvability through semi-supervised learning: a survey of theoretical results[J]. arXiv:1908.09574, 2019.
[5] XU J, HENRIQUES J F, VEDALDI A. Invariant information clustering for unsupervised image classification and segmen-tation[J]. arXiv:1807.06653, 2018.
[6] ZHOU Z H. Machine learning[M]. Beijing: Tsinghua Univer-sity Press, 2018: 311-312.
周志华. 机器学习[M]. 北京: 清华大学出版社, 2018: 311-312.
[7] LIU G H, ZHANG X B. A method for personal identification of communication radiation source based on deep belief net-work[J]. Chinese Journal of Radio Science, 2020, 35(3): 395-403.
刘高辉, 张晓博. 一种基于深度置信网络的通信辐射源个体识别方法[J]. 电波科学学报, 2020, 35(3): 395-403.
[8] HAN S, HAN Q H. Review of semi-supervised learning research[J]. Computer Engineering and Applications, 2020, 56(6): 19-27.
韩嵩, 韩秋弘. 半监督学习研究的述评[J]. 计算机工程与应用, 2020, 56(6): 19-27.
[9] LI H. Statistical learning methods[M]. Beijing: Tsinghua University Press, 2019: 84-88.
李航. 统计学习方法[M]. 北京: 清华大学出版社, 2019: 84-88.
[10] QIU X P. Neural network and deep learning[M]. Beijing: China Machine Press, 2020: 124-129.
邱锡鹏. 神经网络与深度学习[M]. 北京: 机械工业出版社, 2020: 124-129.
[11] CASCANTE-BONILLA P, TAN F W, QI Y J, et al. Curri-culum labeling: self-paced pseudo-labeling for semi-supervised learning[J]. arXiv:2001.06001, 2020.
[12] ZHOU Z H, WANG W, GAO W, et al. Introduction to ma-chine learning theory[M]. Beijing: China Machine Press, 2020.
周志华, 王魏, 高尉, 等. 机器学习理论导引[M]. 北京: 机械工业出版社, 2020.
[13] GRANDVALET Y, BENGIO Y. Semi-supervised learning by entropy minimization[C]//Proceedings of the Advances in Neural Information Processing Systems, Vancouver, Dec 13-18, 2004. Red Hook: Curran Associates, 2004: 529-536.
[14] LEE D. Pseudo-label: the simple and efficient semi-supervised learning method for deep neural networks[C]//Proceedings of the Workshop on Challenges in Representation Learning, Jun 21, 2013: 8-17.
[15] CAI L M, WANG L J. Digital image processing[M]. Beijing: Tsinghua University Press, 2019.
蔡利梅, 王利娟. 数字图像处理[M]. 北京: 清华大学出版社, 2019.
[16] ZHANG X F, WU G. Data augmentation method based on generative adversarial network[J]. Computer Systems & App-lications, 2019, 28(10): 201-206.
张晓峰, 吴刚. 基于生成对抗网络的数据增强方法[J]. 计算机系统应用, 2019, 28(10): 201-206.
[17] CUBUK E D, ZOPH B, MANE D, et al. AutoAugment: learning augmentation strategies from data[C]//Proceedings of the 2019 IEEE Conference on Computer Vision and Pattern Recognition, Long Beach, Jun 16-20, 2019. Piscataway: IEEE, 2019: 113-123.
[18] CUBUK E D, ZOPH B, SHLENS J, et al. RandAugment: practical automated data augmentation with a reduced search space[J]. arXiv:1909.13719, 2019.
[19] ZHANG H, CISSE M, DAUPHIN Y, et al. mixup: beyond-empirical risk minimization[J]. arXiv:1710.09412, 2018.
[20] BLUM A, MITCHELL T M. Combining labeled and unla-beled data with co-training[C]//Proceedings of the 11th Annual Conference on Computational Learning Theory, Madison, Jul 24-26, 1998. New York: ACM, 1998: 92-100.
[21] QIAO S, SHEN W, ZHANG Z S, et al. Deep co-training for semi-supervised image recognition[J]. arXiv:1803.05984, 2018.
[22] CHEN D D, WANG W, GAO W, et al. Tri-net for semi-supervised deep learning[C]//Proceedings of the 27th Inter-national Joint Conference on Artificial Intelligence, Stock-holm, Jul 13-19, 2018: 2014-2020.
[23] CIRESAN D, MEIER U, GAMBARDELLA L, et al. Deep, big, simple neural nets for handwritten digit recognition[J]. Neural Computation, 2010, 22(12): 3207-3220.
[24] LAINE S, AILA T. Temporal ensembling for semisupervised learning[J]. arXiv:1610.02242, 2016.
[25] OUALI Y, HUDELOT C, TAMI M. An overview of deep semi-supervised learning[J]. arXiv:2006.05278, 2020.
[26] TARVAINEN A, VALPOLA H. Mean teachers are better role models: weight-averaged consistency targets improve semi-supervised deep learning results[J]. arXiv:1703.01780, 2018.
[27] XIE Q Z, DAI Z H, HOVY E H, et al. Unsupervised data augmentation for consistency training[J]. arXiv:1904.12848, 2019.
[28] BERTHELOT D, CARLINI N, GOODFELLOW I, et al. MixMatch: a holistic approach to semi-supervised learning[J]. arXiv:1905.02249, 2019.
[29] SOHN K, BERTHELOT D, LI C L, et al. FixMatch: sim-plifying semi-supervised learning with consistency and con-fidence[J]. arXiv:2001.07685, 2020.
[30] GOODFELLOW I J, POUGET-ABADIE J, MIRZA M, et al. Generative adversarial networks[J]. arXiv:1406.2661, 2014.
[31] ZOU X F, ZHU D J. Review on generative adversarial net-work[J]. Computer Systems & Applications, 2019, 28(11): 1-9.
邹秀芳, 朱定局. 生成对抗网络研究综述[J]. 计算机系统应用, 2019, 28(11): 1-9.
[32] SALIMANS T, GOODFELLOW I J, ZAREMBA W, et al. Improved techniques for training GANs[J]. arXiv:1606.03498, 2016.
[33] ODENA A. Semi-supervised learning with generative adver-sarial networks[J]. arXiv:1606.01583, 2016.
[34] KHAN S. Convolution neural network and computer vision[M]. Beijing: China Machine Press, 2019.
KHAN S. 卷积神经网络与计算机视觉[M]. 黄智濒, 戴志涛, 译. 北京: 机械工业出版社, 2019.
[35] KRIZHEVSKY A, HINTON G. Learning multiple layers of features from tiny images[D]. Toronto: University of Toronto, 2009.
[36] COATES A, NG A, LEE H. An analysis of single-layer net-works in unsupervised feature learning[C]//Proceedings of the 14th International Conference on Artificial Intelligence and Statistics, Fort Lauderdale, Apr 16-18, 2011: 215-223.
[37] LU J, GONG P H, YE J P. Learning from very few samples: a survey[J]. arXiv:2009.02653, 2020.
[38] BACHMAN P, HJELM R D, BUCHWALTER W. Learning representations by maximizing mutual information across views[C]//Proceedings of the Annual Conference on Neural Information Processing Systems, Vancouver, Dec 8-14, 2019: 15509-15519.
[39] HJELM R, FEDOROV A, LAVOIE-MARCHILDON S, et al. Learning deep representations by mutual information estimation[J]. arXiv:1808.06670, 2019.
[40] ZHAI X, OLIVER A, KOLESNIKOV A, et al. S4L: self-supervised semi-supervised learning[J]. arXiv:1905.03670, 2019.
[41] SCHMARJE L, SANTAROSSA M, KOCH K. A survey on semi, self and unsupervised learning for image classification[J]. arXiv:2002.08721, 2020.
[42] CHEN J A, YANG Z C, YANG D Y. MixText: linguistically-informed interpolation of hidden space for semi-supervised text classification[J]. arXiv:2004.12239, 2020.
[43] WANG P S, SONG Y, DAI L R. Fine-grained image class-ification with multi-channel visual attention[J]. Journal of Data Acquisition & Processing, 2019, 34(1): 157-166.
王培森, 宋彦, 戴礼荣. 基于多通道视觉注意力的细粒度图像分类[J]. 数据采集与处理, 2019, 34(1): 157-166.
[44] CHENG W J, CHEN W Q. Hyperspectral image classification based on MCFFN-attention[J]. Computer Engineering and Applications, 2020, 56(24): 201-206.
程文娟, 陈文强. 基于MCFFN-Attention的高光谱图像分类[J]. 计算机工程与应用, 2020, 56(24): 201-206.