[1]刘威,王薪予,刘光伟,等.融合关系特征的半监督图像分类方法研究[J].智能系统学报,2022,17(5):886-899.[doi:10.11992/tis.202109022]
　LIU Wei,WANG Xinyu,LIU Guangwei,et al.Semi-supervised image classification method fused with relational features[J].CAAI Transactions on Intelligent Systems,2022,17(5):886-899.[doi:10.11992/tis.202109022]
点击复制

融合关系特征的半监督图像分类方法研究

《智能系统学报》[ISSN 1673-4785/CN 23-1538/TP] 卷: 17 期数: 2022年第5期 页码: 886-899 栏目: 学术论文—机器学习 出版日期: 2022-09-05
Title:
Semi-supervised image classification method fused with relational features
作者:
刘威1,2,3, 王薪予1,3, 刘光伟4, 王东4, 牛英杰1,3
1. 辽宁工程技术大学 理学院，辽宁 阜新 123000;
2. 辽宁工程技术大学 智能工程与数学研究院，辽宁 阜新 123000;
3. 辽宁工程技术大学 数学与系统科学研究所，辽宁 阜新 123000;
4. 辽宁工程技术大学 矿业学院，辽宁 阜新 123000
Author(s):
LIU Wei1,2,3, WANG Xinyu1,3, LIU Guangwei4, WANG Dong4, NIU Yingjie1,3
1. School of Sciences, Liaoning Technical University, Fuxin 123000, China;
2. Institutes of Intelligent Engineering and Mathematics, Liaoning Technical University, Fuxin 123000, China;
3. Institute of Mathematics and Systems Science, Liaoning Technical University, Fuxin 123000, China;
4. School of Mining, Liaoning Technical University, Fuxin 123000, China
关键词:
关系表示; 特征提取; 图卷积神经网络; 融合模型; 半监督学习; 图像分类; 视觉卷积; 泛化性能
Keywords:
relationship representation; feature extraction; graph convolutional neural network; hybrid model; semi-supervised learning; image classification; convolution in vision; generalization performance
分类号:
TP181
DOI:
10.11992/tis.202109022
文献标志码:
2022-06-16
摘要:
半监督深度学习模型具有泛化能力强，所需样本数较少等特点，经过10多年的发展，在理论和实际应用方面都取得了巨大的进步，然而建模样本内部“隐含”关系时模型缺乏解释性以及构造无监督正则化项难度较大等问题限制了半监督深度学习的进一步发展。针对上述问题，从丰富样本特征表示的角度出发，构造了一种新的半监督图像分类模型—融合关系特征的半监督分类模型（semi-supervised classification model fused with relational features，SCUTTLE），该模型在卷积神经网络模型(convolutional neural networks，CNN)基础上引入了图卷积神经网络(graph convolutional networks，GCN)，尝试通过GCN模型来提取CNN模型各层的低、高级特征间的关系，使得融合模型不仅具有特征提取能力，而且具有关系表示能力。通过对SCUTTLE模型泛化性能进行分析，进一步说明了该模型在解决半监督相关问题时的有效性。数值实验结果表明，三层CNN与一层GCN的融合模型在CIFAR10、CIFAR100、SVHN 3种数据集上与CNN监督学习模型的精度相比均可提升5%~6%的精度值，在最先进的ResNet、DenseNet、WRN（wide residual networks）与GCN的融合模型上同样证明了本文所提模型的有效性。
Abstract:
A semi-supervised deep learning model exhibits great generalization ability with minimal required samples and has made great progress in theory and practical application over the past ten years or so. However, the lack of the model’s interpretability when modeling the internal “implicit” relationship of samples and the difficulty in constructing unsupervised regularization items have limited the further development of semi-supervised deep learning. To solve these problems and enrich the sample feature representation, this study has developed a novel semi-supervised model for image classification—semi-supervised classification model integrating the relational features (SCUTTLE). The model introduces the graph convolutional networks (GCN) based on the convolutional neural networks (CNN) and extracts the relationships between the low- and high-level features of each layer of the CNN model via the GCN model, thus extracting features and expressing relationships. By analyzing the generalization performance of the SCUTTLE model, the paper further illustrates its effectiveness in solving semi-supervised related problems. The numerical results indicate that the classification accuracy of the hybrid model with three layers of CNN and one layer of GCN can be improved by 5%–6% compared to that of the CNN model on the CIFAR10, CIFAR100, and SVHN datasets. The effectiveness of the model proposed in this paper is also proved in the most advanced fusion models of ResNet, DenseNet, WRN (wide residual networks), and GCN.
参考文献/References:
[1] 闵帆, 王宏杰, 刘福伦, 等. SUCE: 基于聚类集成的半监督二分类方法[J]. 智能系统学报, 2018, 13(6): 974–980
MIN Fan, WANG Hongjie, LIU Fulun, et al. SUCE: semi-supervised binary classification based on clustering ensemble[J]. CAAI transactions on intelligent systems, 2018, 13(6): 974–980
[2] 姚晓红,黄恒君. 非负半监督函数型聚类方法[J]. 计算机科学与探索, 2021, 15(12): 2438–2448
YAO Xiaohong, HUANG Hengjun. Semi-supervised clustering method for non-negative functional data[J]. Journal of frontiers of computer science and technology, 2021, 15(12): 2438–2448
[3] SAJJADI M, JAVANMARDI M, TASDIZEN T. Regularization with stochastic transformations and perturbations for deep semi-supervised learning[C]//NIPS’16: Proceedings of the 30th International Conference on Neural Information Processing Systems. New York: ACM, 2016: 1171?1179.
[4] CUBUK E D, ZOPH B, MANé D, et al. AutoAugment: learning augmentation strategies from data[C]//2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). Long Beach, IEEE, 2019: 113?123.
[5] LAINE S, AI L A. Temporal ensembling for semi-supervised learning[EB/OL]. (2017?03?15)[2021?01?01]. https://arxiv.org/abs/1610.02242.
[6] LI Yingting, LIU Lu, TAN R T. Certainty-driven consistency loss for semi-supervised learning [EB/OL]. (2017?05?07)[2021?05?01]. https://arxiv.org/abs/1901.05657v1.
[7] ATHIWARATKUN B, FINZI M, IZMAILOV P, et al. There are many consistent explanations of unlabeled data: why You should average [EB/OL]. (2019?02?21)[2021?02?15]. https://arxiv.org/abs/1806.05594.
[8] SZEGEDY C, LIU Wei, JIA Yangqing, et al. Going deeper with convolutions[C]//2015 IEEE Conference on Computer Vision and Pattern Recognition. Boston, MA. IEEE, 2015: 1?9.
[9] CHEN Yunpeng, ROHRBACH M, YAN Zhicheng, et al. Graph-based global reasoning networks[C]//2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). Long Beach, IEEE, 2019: 433?442.
[10] LIU Qinghui, KAMPFFMEYER M, JENSSEN R, et al. Self-constructing graph convolutional networks for semantic labeling[C]//IGARSS 2020 - 2020 IEEE International Geoscience and Remote Sensing Symposium. Waikoloa, IEEE, 2020: 1801?1804.
[11] JIANG Bo, ZHANG Ziyan, LIN Doudou, et al. Semi-supervised learning with graph learning-convolutional networks[C]//2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). Long Beach, IEEE, 2019: 11305?11312.
[12] ZHU? Xiaojin, GHAHRAMANI?H Zoubin. Learning from labeled and unlabeled data with label propagation. CMU-CALD-02-107[R]. Pittsburgh: Carnegie Mellon University, 2002.
[13] WESTON J, RATLE F, MOBAHI H, et al. Deep learning via semi-supervised embedding[M]//Lecture Notes in Computer Science. Berlin, Heidelberg: Springer Berlin Heidelberg, 2012: 639?655.
[14] GOODFELLOW I, POUGET-ABADIE J, MIRZA M, et al. Generative adversarial nets[J]. Advances in neural information processing systems, 2014, 27(1): 2672–2680.
[15] SPRINGENBERG J T. Unsupervised and semi-supervised learning with categorical generative adversarial networks[EB/OL].(2016?04?30)[2020?05?15].https://arxiv.org/abs/1511.06390.
[16] CHANG Chuanyu, CHEN T Y, CHUNG P C. Semi-supervised learning using generative adversarial networks[C]//2018 IEEE Symposium Series on Computational Intelligence. Bangalore, IEEE, 2018: 892?896.
[17] BELKIN M, NIYOGI P. Laplacian eigenmaps for dimensionality reduction and data representation[J]. Neural computation, 2003, 15(6): 1373–1396.
[18] NAIR V, HINTON G E. Rectified linear units improve restricted boltzmann machines[C]//Proceedings of the ICML. Haifa, Israel, 2010: 807?814.
[19] KIPF T N, WELLING M. Semi-supervised classification with graph convolutional networks[EB/OL]. (2017?02?22)[2019?12?15]. https://arxiv.org/abs/1609.02907.
[20] MOORE A W. An intoductory tutorial on kd-trees. Efficient Memory-based Learning for Robot Control - 209[R]. Computer Laboratory, University of Cambridge. 1991.
[21] GOODFELLOW I J, BULATOV Y, IBARZ J, et al. Multi-digit number recognition from street view imagery using deep convolutional neural networks[EB/OL]. (2014?04?14)[2019?06?05]. https://arxiv.org/abs/1312.6082.
[22] SIMONYAN K, ZISSERMAN A. Very deep convolutional networks for large-scale image recognition[EB/OL]. (2015?04?10)[2020?04?20]. https://arxiv.org/abs/1409.1556
[23] ZAGORUYKO S, KOMODAKIS N. Wide residual networks[EB/OL]. (2017?06?14)[2020?12?15]. https://arxiv.org/abs/1605.07146.
[24] OLIVER A, ODENA A, RAFFEL C A, et al. Realistic evaluation of deep semi-supervised learning algorithms[J]. Advances in neural information processing systems, 2018, 31(1): 3235–3246.
[25] TARVAINEN A, VALPOLA H. Mean teachers are better role models: Weight-averaged consistency targets improve semi-supervised deep learning results[J]. Advances in neural information processing systems, 2017, 30(1): 1195–1204.
[26] MIYATO T, MAEDA S I, KOYAMA M, et al. Virtual adversarial training: a regularization method for supervised and semi-supervised learning[J]. IEEE transactions on pattern analysis and machine intelligence, 2019, 41(8): 1979–1993.
[27] ZHANG Hongyi, CISSE M, DAUPHIN Y N, et al. mixup: Beyond empirical risk minimization[EB/OL].(2018?04?27)[2020?04?01]. https://arxiv.org/abs/1710.09412.
[28] LEE D H. Pseudo-label: The simple and efficient semi-supervised learning method for deep neural networks[C]//Workshop on challenges in representation learning, ICML. 2013, 3(2): 896?901.
相似文献/References:
[1]黄剑华,唐降龙,刘家锋,等.一种基于Homogeneity的文本检测新方法[J].智能系统学报,2007,2(1):69.
　HUANG Jian-hua,TANG Xiang-long,LIU Jia-feng,et al.A new method for text detection based on Homogeneity[J].CAAI Transactions on Intelligent Systems,2007,2():69.
[2]谭 营,朱元春.反垃圾电子邮件方法研究进展[J].智能系统学报,2010,5(3):189.
　TAN Ying,ZHU Yuan-chun.Advances in antispam techniques[J].CAAI Transactions on Intelligent Systems,2010,5():189.
[3]王斐,张育中,宁廷会,等.脑-机接口研究进展[J].智能系统学报,2011,6(3):189.
　WANG Fei,ZHANG Yuzhong,NING Tinghui,et al.Research progress in a braincomputer interface[J].CAAI Transactions on Intelligent Systems,2011,6():189.
[4]刘琚,孙建德.独立分量分析的图像/视频分析与应用[J].智能系统学报,2011,6(6):495.
　LIU Ju,SUN Jiande.Independent component analysisbased image/video analysis and applications[J].CAAI Transactions on Intelligent Systems,2011,6():495.
[5]谭营,王军.手指静脉身份识别技术最新进展[J].智能系统学报,2011,6(6):471.
　TAN Ying,WANG Jun.Recent advances in finger vein based biometric techniques[J].CAAI Transactions on Intelligent Systems,2011,6():471.
[6]吴家伟,严京旗,方志宏,等.基于图像显著性特征的铸坯表面缺陷检测[J].智能系统学报,2012,7(1):75.
　WU Jiawei,YAN Jingqi,FANG Zhihong,et al.Defect detection on a steel slab surface based on the characteristics of an image’s saliency region[J].CAAI Transactions on Intelligent Systems,2012,7():75.
[7]张毅,罗明伟,罗元.脑电信号的小波变换和样本熵特征提取方法[J].智能系统学报,2012,7(4):339.
　ZHANG Yi,LUO Mingwei,LUO Yuan.EEG feature extraction method based on wavelet transform and sample entropy[J].CAAI Transactions on Intelligent Systems,2012,7():339.
[8]刘忠宝,王士同.从Parzen窗核密度估计到特征提取方法：新的研究视角[J].智能系统学报,2012,7(6):471.
　LIU Zhongbao,WANG Shitong.From Parzen window estimation to feature extraction: a new perspective[J].CAAI Transactions on Intelligent Systems,2012,7():471.
[9]孙倩茹,王文敏,刘宏.视频序列的人体运动描述方法综述[J].智能系统学报,2013,8(3):189.
　SUN Qianru,WANG Wenmin,LIU Hong.Study of human action representation in video sequences[J].CAAI Transactions on Intelligent Systems,2013,8():189.
[10]许可乐,唐涛,蒋咏梅.一种SAR图像稳健特征点提取方法[J].智能系统学报,2013,8(4):287.[doi:10.3969/j.issn.1673-4785.201304038]
　XU Kele,TANG Tao,JIANG Yongmei.A stable feature point extraction approach for SAR image registration[J].CAAI Transactions on Intelligent Systems,2013,8():287.[doi:10.3969/j.issn.1673-4785.201304038]

备注/Memo

收稿日期:2021-09-13。
基金项目:国家自然科学基金项目(51974144, 51874160)；辽宁省教育厅项目(LJKZ0340)；辽宁工程技术大学学科创新团队项目(LNTU20TD-01,LNTU20TD- 07).
作者简介:刘威，副教授，博士，主要研究方向为深度神经网络、机器学习、矿业系统工程。获得中国煤炭工业科学技术奖二等奖、三等奖等奖项，发表学术论文30余篇，参与编写学术专著及教材3部。;王薪予，硕士研究生，主要研究方向为机器学习与深度学习、图神经网络;刘光伟，教授，博士生导师，博士，主要研究方向为露天矿开采设计理论、矿业系统工程
通讯作者:刘威. E-mail：lv8218218@126.com

更新日期/Last Update: 1900-01-01
Copyright © 《 智能系统学报》 编辑部
地址：（150001）黑龙江省哈尔滨市南岗区南通大街145-1号楼 电话：0451- 82534001、82518134　邮箱：tis@vip.sina.com