基于元学习的标签噪声自适应学习算法

doi:10.11772/j.issn.1001-9081.2024070932

《计算机应用》唯一官方网站 ›› 2025, Vol. 45 ›› Issue (7): 2113-2122.DOI: 10.11772/j.issn.1001-9081.2024070932

• CCF第39届中国计算机应用大会 (CCF NCCA 2024) • 上一篇下一篇

基于元学习的标签噪声自适应学习算法

齐巧玲¹, 王啸啸², 张茜茜³(), 汪鹏², 董永峰²^,⁴^,⁵

^1.河北工业大学电气工程学院，天津 300130
^2.河北工业大学人工智能与数据科学学院，天津 300401
^3.河北工业大学能源与环境工程学院，天津 300401
^4.天津市虚拟现实与可视计算国际联合中心，天津 300401
^5.河北省数据驱动工业智能工程研究中心（河北工业大学），天津 300401

收稿日期:2024-07-05 修回日期:2024-10-15 接受日期:2024-10-16 发布日期:2025-07-10 出版日期:2025-07-10
通讯作者: 张茜茜
作者简介:齐巧玲（1984—），女，河北晋州人，讲师，硕士，主要研究方向：人工智能
王啸啸（1998—），女，河北衡水人，硕士研究生，主要研究方向：图像处理
汪鹏（1978—），男，河北邯郸人，副教授，博士，CCF会员，主要研究方向：机器学习、智能计算
董永峰（1977—），男，河北定州人，教授，博士，CCF会员，主要研究方向：计算机视觉、智能信息处理。
基金资助:
中国高等教育学会高等教育科学研究规划课题(22XX0401);河北省首批省级研究生教育教学改革研究项目(YJG2023024)

Label noise adaptive learning algorithm based on meta-learning

Qiaoling QI¹, Xiaoxiao WANG², Qianqian ZHANG³(), Peng WANG², Yongfeng DONG²^,⁴^,⁵

^1.School of Electrical Engineering，Hebei University of Technology，Tianjin 300130，China
^2.School of Artificial Intelligence，Hebei University of Technology，Tianjin 300401，China
^3.School of Energy and Environmental Engineering，Hebei University of Technology，Tianjin 300401，China
^4.Tianjin International Joint Center for Virtual Reality and Visual Computing，Tianjin 300401，China
^5.Hebei Data Driven Industrial Intelligent Engineering Research Center （Hebei University of Technology），Tianjin 300401，China

Received:2024-07-05 Revised:2024-10-15 Accepted:2024-10-16 Online:2025-07-10 Published:2025-07-10
Contact: Qianqian ZHANG
About author:QI Qiaoling， born in 1984， M. S.， lecturer. Her research interests include artificial intelligence.
WANG Xiaoxiao， born in 1998， M. S. candidate. Her research interests include image processing.
WANG Peng， born in 1978， Ph. D.， associate professor. His research interests include machine learning， intelligent computing.
DONG Yongfeng， born in 1977， Ph. D.， professor. His research interests include computer vision， intelligent information processing.
Supported by:
Higher Education Scientific Research Planning of China Association of Higher Education Project(22XX0401);First Batch of Provincial Postgraduate Education and Teaching Reform Research Project in Hebei Province(YJG2023024)

摘要/Abstract

摘要：

图像分类需要收集大量的图片进行模型训练与优化，但收集过程会不可避免地带来噪声标签。为了应对这一挑战，鲁棒性分类方法应运而生。在目前的鲁棒性分类方法中，超参数的设置需要手动调节，对人力物力带来了大量的损耗。因此，提出元超参数调节器（MHA），采用双层嵌套循环优化的方法自适应地学习噪声感知的超参数组合，并提出Meta-FPL （Feature Pseudo-Label adaptive learning based on Meta learning）算法。此外，为了解决元训练阶段的反向传播过程耗费大量GPU算力的问题，提出选择激活元模型层（SAML）策略。该策略通过比较虚拟训练阶段反向传播的平均梯度与元梯度的大小，限制部分元模型层的更新，从而有效提升模型的训练效率。在4个基准数据集和1个真实数据集上的实验结果表明，与MLC（Meta Label Correction for noisy label learning）、CTRR（ConTrastive RegulaRization）和FPL（Feature Pseudo-Label）算法相比，Meta-FPL算法的分类准确率较高。此外，引入SAML策略后，在元训练阶段的反向传播过程训练时长缩短了79.52%。可见，Meta-FPL算法能在较短的训练时间内有效提升分类测试准确率。

关键词: 深度学习, 深度神经网络, 图像分类, 标签噪声, 元学习

Abstract:

Image classification requires the collection of a large number of images for model training and optimization， but the image collection process will bring noisy labels inevitably. To cope with this challenge， robust classification methods have emerged. The setting of hyperparameters in the current robust classification methods needs to be adjusted manually， which brings a lot of loss in human and material resources. Therefore， Meta Hyperparameter Adjuster （MHA） was proposed， which adopted a two-layer nested loop optimization method to learn noise-aware hyperparameter combinations adaptively， and a Meta-FPL （Feature Pseudo-Label adaptive learning algorithm based on Meta learning） algorithm was proposed too. In addition， in order to solve the problem that the backpropagation process in meta training phase consumes a large amount of GPU computing power， the Select Activation Metamodel Layer （SAML） strategy was proposed， which restricts the update of some metamodel layers by comparing sizes of the average gradient of the backpropagation and the meta-gradient in virtual training phase， which improves training efficiency of the model effectively. Experimental results on four benchmark datasets and one real dataset show that compared with MLC （Meta Label Correction for noisy label learning）， CTRR （ConTrastive RegulaRization） and Feature Pseudo Label （FPL） algorithms， Meta-FPL algorithm has a higher classification accuracy. In addition， after introducing SAML strategy， the training duration of the backpropagation process in the meta training phase was reduced by 79.52%. It can be seen that Meta-FPL algorithm can effectively improve the accuracy of classification testing in a shorter training time.

Key words: deep learning, deep neural network, image classification, label noise, meta-learning

中图分类号:

TP391.4

齐巧玲, 王啸啸, 张茜茜, 汪鹏, 董永峰. 基于元学习的标签噪声自适应学习算法[J]. 计算机应用, 2025, 45(7): 2113-2122.

Qiaoling QI, Xiaoxiao WANG, Qianqian ZHANG, Peng WANG, Yongfeng DONG. Label noise adaptive learning algorithm based on meta-learning[J]. Journal of Computer Applications, 2025, 45(7): 2113-2122.

图/表 13

参考文献 48

[1]	JORDAN M I， MITCHELL T M. Machine learning： trends， perspectives， and prospects ［J］. Science， 2015， 349（6245）： 255-260.
[2]	WANG X， XIE D， ZHENG Y. Referring expression grounding by multi-context reasoning ［J］. Pattern Recognition Letters， 2022， 160： 66-72.
[3]	KRIZHEVSKY A， SUTSKEVER I， HINTON G E. ImageNet classification with deep convolutional neural networks ［J］. Communications of the ACM， 2017， 60（6）： 84-90.
[4]	GIRSHICK R， DONAHUE J， DARRELL T， et al. Rich feature hierarchies for accurate object detection and semantic segmentation ［C］// Proceedings of the 2014 IEEE Conference on Computer Vision and Pattern Recognition. Piscataway： IEEE， 2014： 580-587.
[5]	LIN T Y， DOLLÁR P， GIRSHICK R， et al. Feature pyramid networks for object detection ［C］// Proceedings of the 2017 IEEE Conference on Computer Vision and Pattern Recognition. Piscataway： IEEE， 2017： 936-944.
[6]	CHEN L C， PAPANDREOU G， KOKKINOS I， et al. DeepLab： Semantic image segmentation with deep convolutional nets， atrous convolution， and fully connected CRFs ［J］. IEEE Transactions on Pattern Analysis and Machine Intelligence， 2018， 40（4）： 834-848.
[7]	HU G， HUA Y， YUAN Y， et al. Attribute-enhanced face recognition with neural tensor fusion networks ［C］// Proceedings of the 2017 IEEE International Conference on Computer Vision. Piscataway： IEEE， 2017： 3764-3773.
[8]	LEE D H. Pseudo-label： the simple and efficient semi-supervised learning method for deep neural networks ［EB/OL］. ［2024-02-16］. .
[9]	黄涵.基于深度学习的标签噪声学习算法综述［J］.信息系统工程，2023（10）： 39-42.
	HUANG H. A review of label noise learning algorithms based on deep learning ［J］. China CIO News， 2023（10）： 39-42.
[10]	FRÉNAY B， VERLEYSEN M. Classification in the presence of label noise： a survey ［J］. IEEE Transactions on Neural Networks and Learning Systems， 2014， 25（5）： 845-869.
[11]	FRÉNAY B， KABÁN A. A comprehensive introduction to label noise ［EB/OL］. ［2024-02-16］. .
[12]	ALGAN G， ULUSOY I. Image classification with deep learning in the presence of noisy labels： a survey ［J］. Knowledge-Based Systems， 2021， 215： No.106771.
[13]	ARPIT D， JASTRZĘBSKI S， BALLAS N， et al. A closer look at memorization in deep networks ［C］// Proceedings of the 34th International Conference on Machine Learning. New York： JMLR.org， 2017： 233-242.
[14]	SONG H， KIM M， PARK D， et al. Learning from noisy labels with deep neural networks： a survey ［J］. IEEE Transactions on Neural Networks and Learning Systems， 2023， 34（11）： 8135-8153.
[15]	BARRON J T. General and adaptive robust loss function ［C］// Proceedings of the 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Piscataway： IEEE， 2019： 4326-4334.
[16]	WANG Y， LIU W， MA X， et al. Iterative learning with open-set noisy labels ［C］// Proceedings of the 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Piscataway： IEEE， 2018： 8688-8696.
[17]	ZHANG Z， SABUNCU M R. Generalized cross entropy loss for training deep neural networks with noisy labels ［C］// Proceedings of the 32nd International Conference on Neural Information Processing Systems. Red Hook： Curran Associates Inc.， 2018： 8792-8802.
[18]	WANG Y， MA X， CHEN Z， et al. Symmetric cross entropy for robust learning with noisy labels ［C］// Proceedings of the 2019 IEEE/CVF International Conference on Computer Vision. Piscataway： IEEE， 2019： 322-330.
[19]	FENG L， SHU S， LIN Z， et al. Can cross entropy loss be robust to label noise？［C］// Proceedings of the 29th International Joint Conferences on Artificial Intelligence. California： ijcai.org， 2021： 2206-2212.
[20]	MA X， HUANG H， WANG Y， et al. Normalized loss functions for deep learning with noisy labels ［C］// Proceedings of the 37th International Conference on Machine Learning. New York： JMLR.org， 2020： 6543-6553.
[21]	SHU J， ZHAO Q， CHEN K， et al. Learning adaptive loss for robust learning with noisy labels ［EB/OL］. ［2024-02-16］. .
[22]	DONG Y， LI J， WANG Z， et al. CoDC： accurate learning with noisy labels via disagreement and consistency ［J］. Biomimetics， 2024， 9（2）： No.92.
[23]	PATRINI G， ROZZA A， MENON A K， et al. Making deep neural networks robust to label noise： a loss correction approach ［C］// Proceedings of the 2017 IEEE Conference on Computer Vision and Pattern Recognition. Piscataway： IEEE， 2017： 2233-2241.
[24]	XIA X， LIU T， WANG N， et al. Are anchor points really indispensable in label-noise learning？［C］// Proceedings of the 33rd International Conference on Neural Information Processing Systems. Red Hook： Curran Associates Inc.， 2019： 6838-6849.
[25]	WANG Z， HU G， HU Q. Training noise-robust deep neural networks via meta-learning ［C］// Proceedings of the 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Piscataway： IEEE， 2020： 4523-4532.
[26]	KANG B， XIE S， ROHRBACH M， et al. Decoupling representation and classifier for long-tailed recognition ［EB/OL］. ［2024-02-16］. .
[27]	LIN T Y， GOYAL P， GIRSHICK R， et al. Focal loss for dense object detection ［C］// Proceedings of the 2017 IEEE Conference on Computer Vision and Pattern Recognition. Piscataway： IEEE， 2017： 2999-3007.
[28]	HAN B， YAO Q， YU X， et al. Co-teaching： robust training of deep neural networks with extremely noisy labels ［C］// Proceedings of the 32nd International Conference on Neural Information Processing Systems. Red Hook： Curran Associates Inc.， 2018： 8536-8546.
[29]	SHU J， XIE Q， YI L， et al. Meta-weight-net： learning an explicit mapping for sample weighting ［C］// Proceedings of the 33rd International Conference on Neural Information Processing Systems. Red Hook： Curran Associates Inc.， 2019： 1919-1930.
[30]	HU Z， YANG Z， HU X， et al. SimPLE： similar pseudo label exploitation for semi-supervised classification ［C］// Proceedings of the 2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Piscataway： IEEE， 2021： 15094-15103.
[31]	KARIM N， RIZVE M N， RAHNAVARD N， et al. UNICON： combating label noise through uniform selection and contrastive learning ［C］// Proceedings of the 2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Piscataway： IEEE， 2022： 9666-9676.
[32]	SONG H， KIM M， LEE J G. SELFIE： refurbishing unclean samples for robust deep learning ［C］// Proceedings of the 36th International Conference on Machine Learning. New York： JMLR.org， 2019： 5907-5915.
[33]	PHAM H， DAI Z， XIE Q， et al. Meta pseudo labels ［C］// Proceedings of the 2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Piscataway： IEEE， 2021： 11552-11563.
[34]	WU Y， SHU J， XIE Q， et al. Learning to purify noisy labels via meta soft label corrector ［C］// Proceedings of the 35th AAAI Conference on Artificial Intelligence. Palo Alto： AAAI Press， 2021： 10388-10396.
[35]	ZHENG G， AWADALLAH A H， DUMAIS S. Meta label correction for noisy label learning ［C］// Proceedings of the 35th AAAI Conference on Artificial Intelligence. Palo Alto： AAAI Press， 2021： 11053-11061.
[36]	WANG P， WANG X， WANG Z， et al. Learning accurate pseudo-labels via feature similarity in the presence of label noise ［J］. Applied Sciences， 2024， 14（7）： No.2759.
[37]	PLEISS G， ZHANG T， ELENBERG E， et al. Identifying mislabeled data using the area under the margin ranking ［C］// Proceedings of the 34th International Conference on Neural Information Processing Systems. Red Hook： Curran Associates Inc.， 2020： 17044-17056.
[38]	DENG L. The MNIST database of handwritten digit images for machine learning research ［Best of the Web］［J］. IEEE Signal Processing Magazine， 2012， 29（6）： 141-142.
[39]	NETZER Y， WANG T， COATES A， et al. Reading digits in natural images with unsupervised feature learning ［EB/OL］. ［2024-02-16］. .
[40]	KRIZHEVSKY A. Learning multiple layers of features from tiny images ［R/OL］. ［2024-02-16］. .
[41]	TANAKA D， IKAMI D， YAMASAKI T， et al. Joint optimization framework for learning with noisy labels ［C］// Proceedings of the 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Piscataway： IEEE， 2018： 5552-5560.
[42]	LARSEN J， NONBOE L， HINTZ-MADSEN M， et al. Design of robust neural network classifiers ［C］// Proceedings of the 1998 International Conference on Acoustics， Speech and Signal Processing — Volume 2. Piscataway： IEEE， 1998： 1205-1208.
[43]	ZHAO G， LI G， QIN Y， et al. Centrality and consistency： two-stage clean samples identification for learning with instance-dependent noisy labels ［C］// Proceedings of the 2022 European Conference on Computer Vision， LNCS 13685. Cham： Springer， 2022： 21-37.
[44]	LIU S， NILES-WEED J， RAZAVIAN N， et al. Early-learning regularization prevents memorization of noisy labels ［C］// Proceedings of the 34th International Conference on Neural Information Processing Systems. Red Hook： Curran Associates Inc.， 2020： 20331-20342.
[45]	XIA X， LIU T， HAN B， et al. Robust early-learning： hindering the memorization of noisy labels ［EB/OL］. ［2024-02-16］. .
[46]	BAI Y， YANG E， HAN B， et al. Understanding and improving early stopping for learning with noisy labels ［C］// Proceedings of the 35th International Conference on Neural Information Processing Systems. Red Hook： Curran Associates Inc.， 2021： 24392-24403.
[47]	YI L， LIU S， SHE Q， et al. On learning contrastive representations for learning with noisy labels ［C］// Proceedings of the 2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Piscataway： IEEE， 2022： 16661-16670.
[48]	MILLER A C， FOTI N J， D’AMOUR A， et al. Reducing reparameterization gradient variance ［C］// Proceedings of the 31st International Conference on Neural Information Processing Systems. Red Hook： Curran Associates Inc.， 2017： 3711-3721.

数据集	类别数	训练集样本数	测试集样本数	图片尺寸
MNIST	10	60 000	10 000	28×28
SVHN	10	7 325	26 032	32×32
CIFAR10	10	50 000	10 000	32×32
CIFAR100	100	50 000	10 000	32×32
Clothing1M	14	50 000	14 000	224×224

数据集	类别数	训练集样本数	测试集样本数	图片尺寸
MNIST	10	60 000	10 000	28×28
SVHN	10	7 325	26 032	32×32
CIFAR10	10	50 000	10 000	32×32
CIFAR100	100	50 000	10 000	32×32
Clothing1M	14	50 000	14 000	224×224

算法	Symmetric		Pairflip		Tridiagonal		Instance
算法	NR为20%	NR为40%	NR为20%	NR为40%	NR为20%	NR为40%	NR为20%	NR为40%
Standard	94.52±0.16	78.56±0.36	88.93±0.38	64.94±0.42	91.70±0.20	71.26±0.31	90.15±0.25	69.24±0.41
GCE	98.99±0.14	98.46±0.12	99.16±0.11	98.75±0.13	99.12±0.09	99.03±0.14	98.28±0.17	97.97±0.21
APL	99.53±0.03	99.29±0.03	99.22±0.05	80.01±0.43	99.41±0.04	97.22±0.04	99.26±0.04	87.47±0.44
ELR	99.22±0.13	98.97±0.16	99.09±0.10	98.99±0.14	99.09±0.09	99.06±0.15	99.05±0.11	98.94±0.17
MSLC	99.07±0.11	98.85±0.14	99.57±0.10	99.33±0.10	99.64±0.12	99.35±0.14	99.58±0.14	99.41±0.16
CDR	94.27±0.35	75.76±0.94	87.40±0.94	62.83±1.57	91.40±0.45	70.13±1.16	89.62±0.53	67.40±1.59
PES	99.57±0.04	99.49±0.05	99.62±0.02	99.58±0.03	99.69±0.01	99.54±0.03	99.51±0.02	99.42±0.04
MLC	98.80±0.04	98.47±0.07	98.94±0.05	97.36±0.06	98.91±0.01	98.54±0.03	98.94±0.03	98.42±0.05
CTRR	98.96±0.24	98.16±0.34	99.29±0.52	99.06±0.47	99.01±0.35	98.94±0.26	98.83±0.28	98.14±0.32
FPL	99.71±0.01	99.71±0.02	99.72±0.02	99.70±0.02	99.73±0.02	99.70±0.02	99.60±0.01	99.52±0.02
Meta-FPL	99.80±0.03	99.75±0.05	99.83±0.04	99.77±0.03	99.79±0.04	99.72±0.04	99.71±0.04	99.64±0.05

算法	Symmetric		Pairflip		Tridiagonal		Instance
算法	NR为20%	NR为40%	NR为20%	NR为40%	NR为20%	NR为40%	NR为20%	NR为40%
Standard	94.52±0.16	78.56±0.36	88.93±0.38	64.94±0.42	91.70±0.20	71.26±0.31	90.15±0.25	69.24±0.41
GCE	98.99±0.14	98.46±0.12	99.16±0.11	98.75±0.13	99.12±0.09	99.03±0.14	98.28±0.17	97.97±0.21
APL	99.53±0.03	99.29±0.03	99.22±0.05	80.01±0.43	99.41±0.04	97.22±0.04	99.26±0.04	87.47±0.44
ELR	99.22±0.13	98.97±0.16	99.09±0.10	98.99±0.14	99.09±0.09	99.06±0.15	99.05±0.11	98.94±0.17
MSLC	99.07±0.11	98.85±0.14	99.57±0.10	99.33±0.10	99.64±0.12	99.35±0.14	99.58±0.14	99.41±0.16
CDR	94.27±0.35	75.76±0.94	87.40±0.94	62.83±1.57	91.40±0.45	70.13±1.16	89.62±0.53	67.40±1.59
PES	99.57±0.04	99.49±0.05	99.62±0.02	99.58±0.03	99.69±0.01	99.54±0.03	99.51±0.02	99.42±0.04
MLC	98.80±0.04	98.47±0.07	98.94±0.05	97.36±0.06	98.91±0.01	98.54±0.03	98.94±0.03	98.42±0.05
CTRR	98.96±0.24	98.16±0.34	99.29±0.52	99.06±0.47	99.01±0.35	98.94±0.26	98.83±0.28	98.14±0.32
FPL	99.71±0.01	99.71±0.02	99.72±0.02	99.70±0.02	99.73±0.02	99.70±0.02	99.60±0.01	99.52±0.02
Meta-FPL	99.80±0.03	99.75±0.05	99.83±0.04	99.77±0.03	99.79±0.04	99.72±0.04	99.71±0.04	99.64±0.05

算法	Symmetric		Pairflip		Tridiagonal		Instance
算法	NR为20%	NR为40%	NR为20%	NR为40%	NR为20%	NR为40%	NR为20%	NR为40%
Standard	87.44±0.34	67.09±0.56	83.42±0.43	60.86±0.37	86.0±0.24	67.21±0.44	83.2±0.45	59.55±0.64
GCE	76.57±0.28	60.82±0.30	77.83±0.24	62.59±0.35	78.06±0.19	62.98±0.27	75.87±0.26	60.12±0.34
APL	77.46±0.31	61.05±0.46	75.62±0.45	59.96±0.47	76.06±0.34	60.53±0.41	75.27±0.44	59.83±0.45
ELR	74.39±0.28	56.19±0.23	76.94±0.22	57.43±0.26	75.94±0.28	57.16±0.21	75.69±0.32	58.16±0.27
MSLC	97.08±0.17	95.12±0.14	96.97±0.21	95.24±0.23	97.10±0.18	96.88±0.20	97.09±0.23	96.95±0.25
CDR	82.46±0.21	67.50±2.09	82.35±1.39	60.41±2.15	84.79±1.00	64.59±1.44	81.96±2.30	60.71±3.33
PES	87.63±0.23	69.31±0.37	88.92±0.25	70.26±0.31	89.17±0.19	70.34±0.27	85.49±0.21	67.38±0.36
MLC	96.75±0.06	95.96±0.12	96.99±0.11	96.06±0.10	96.43±0.09	95.81±0.11	96.38±0.12	95.24±0.15
CTRR	96.21±0.14	95.57±0.53	97.02±0.21	96.20±0.34	96.53±0.44	96.06±0.45	96.17±0.14	95.89±0.11
FPL	97.14±0.02	96.87±0.02	97.15±0.01	97.11±0.02	97.14±0.01	97.01±0.01	97.11±0.01	97.02±0.01
Meta-FPL	97.23±0.04	96.98±0.05	97.21±0.03	97.14±0.04	97.26±0.05	97.20±0.07	97.16±0.06	97.09±0.06

基于元学习的标签噪声自适应学习算法

Label noise adaptive learning algorithm based on meta-learning

RichHTML

PDF

可视化

摘要/Abstract

引用本文

使用本文

图/表 13

参考文献 48

相关文章 15

编辑推荐

Metrics

算法	最终	最佳
Standard	67.22	64.68
APL	56.01	55.84
ELR	73.27	74.53
MSLC	73.34	73.95
CDR	63.72	63.22
PES	72.37	73.64
MLC	73.28	75.61
CTRR	72.71	73.68
FPL	73.54	73.57
Meta-FPL	73.61	73.94

[1]	王静, 刘嘉星, 宋婉莹, 薛嘉兴, 丁温欣. 基于空间变换网络和特征分布校准的小样本皮肤图像分类模型[J]. 《计算机应用》唯一官方网站, 2025, 45(8): 2720-2726.
[2]	廖炎华, 鄢元霞, 潘文林. 基于YOLOv9的交通路口图像的多目标检测算法[J]. 《计算机应用》唯一官方网站, 2025, 45(8): 2555-2565.
[3]	彭鹏, 蔡子婷, 刘雯玲, 陈才华, 曾维, 黄宝来. 基于CNN和双向GRU混合孪生网络的语音情感识别方法[J]. 《计算机应用》唯一官方网站, 2025, 45(8): 2515-2521.
[4]	张硕, 孙国凯, 庄园, 冯小雨, 王敬之. 面向区块链节点分析的eclipse攻击动态检测方法[J]. 《计算机应用》唯一官方网站, 2025, 45(8): 2428-2436.
[5]	葛丽娜, 王明禹, 田蕾. 联邦学习的高效性研究综述[J]. 《计算机应用》唯一官方网站, 2025, 45(8): 2387-2398.
[6]	张子墨, 赵雪专. 多尺度稀疏图引导的视觉图神经网络[J]. 《计算机应用》唯一官方网站, 2025, 45(7): 2188-2194.
[7]	向尔康, 黄荣, 董爱华. 开放生成与特征优化的开集识别方法[J]. 《计算机应用》唯一官方网站, 2025, 45(7): 2195-2202.
[8]	索晋贤, 张丽萍, 闫盛, 王东奇, 张雅雯. 可解释的深度知识追踪方法综述[J]. 《计算机应用》唯一官方网站, 2025, 45(7): 2043-2055.
[9]	王震洲, 郭方方, 宿景芳, 苏鹤, 王建超. 面向智能巡检的视觉模型鲁棒性优化方法[J]. 《计算机应用》唯一官方网站, 2025, 45(7): 2361-2368.
[10]	王慧斌, 胡展傲, 胡节, 徐袁伟, 文博. 基于分段注意力机制的时间序列预测模型[J]. 《计算机应用》唯一官方网站, 2025, 45(7): 2262-2268.
[11]	赵小阳, 许新征, 李仲年. 物联网应用中的可解释人工智能研究综述[J]. 《计算机应用》唯一官方网站, 2025, 45(7): 2169-2179.
[12]	花天辰, 马晓宁, 智慧. 基于浅层人工神经网络的可移植执行恶意软件静态检测模型[J]. 《计算机应用》唯一官方网站, 2025, 45(6): 1911-1921.
[13]	李岚皓, 严皓钧, 周号益, 孙庆赟, 李建欣. 基于神经网络的多尺度信息融合时间序列长期预测模型[J]. 《计算机应用》唯一官方网站, 2025, 45(6): 1776-1783.
[14]	王向, 崔倩倩, 张晓明, 王建超, 王震洲, 宋佳霖. 改进ConvNeXt的无线胶囊内镜图像分类模型[J]. 《计算机应用》唯一官方网站, 2025, 45(6): 2016-2024.
[15]	王文鹏, 秦寅畅, 师文轩. 工业缺陷检测无监督深度学习方法综述[J]. 《计算机应用》唯一官方网站, 2025, 45(5): 1658-1670.

算法	Symmetric		Pairflip		Tridiagonal		Instance
算法	NR为20%	NR为40%	NR为20%	NR为40%	NR为20%	NR为40%	NR为20%	NR为40%
Standard	83.81±0.15	66.22±0.43	81.83±0.32	59.41±0.60	82.97±0.23	64.18±0.41	81.65±0.42	59.02±0.42
GCE	89.70±0.23	87.62±0.35	89.83±0.19	87.61±0.48	90.35±0.04	86.63±0.06	89.45±0.29	57.41±0.34
APL	88.15±0.09	81.51±0.17	86.07±0.09	51.66±0.20	77.34±0.23	32.66±0.26	78.83±2.11	60.00±1.71
ELR	89.42±0.11	87.52±0.10	89.87±0.10	89.36±0.16	88.96±0.09	88.48±0.14	89.78±0.13	90.05±0.17
MSLC	93.20±0.09	90.67±0.17	94.09±0.09	92.74±0.11	92.82±0.11	92.44±0.11	92.46±0.10	91.93±0.14
CDR	82.79±0.66	64.16±0.74	81.25±0.96	60.17±0.94	82.14±0.60	64.13±1.13	81.11±0.80	62.69±1.51
PES	92.48±0.06	89.17±0.13	92.14±0.03	88.77±0.09	92.40±0.05	88.70±0.08	91.57±0.07	89.27±0.12
MLC	89.46±0.12	86.53±0.24	90.33±0.15	90.30±0.14	90.31±0.15	90.29±0.13	89.74±0.13	89.51±0.14
CTRR	92.97±0.32	92.16±0.31	93.05±0.56	92.82±0.27	93.89±0.39	93.04±0.33	91.67±0.25	90.16±0.50
FPL	93.55±0.03	92.92±0.04	94.13±0.07	94.08±0.09	94.23±0.06	93.71±0.06	93.70±0.06	93.50±0.04
Meta-FPL	93.60±0.05	93.01±0.05	94.20±0.06	94.12±0.07	94.32±0.07	93.78±0.06	93.77±0.05	93.60±0.08

算法	Symmetric		Pairflip		Tridiagonal		Instance
算法	NR为20%	NR为40%	NR为20%	NR为40%	NR为20%	NR为40%	NR为20%	NR为40%
Standard	83.81±0.15	66.22±0.43	81.83±0.32	59.41±0.60	82.97±0.23	64.18±0.41	81.65±0.42	59.02±0.42
GCE	89.70±0.23	87.62±0.35	89.83±0.19	87.61±0.48	90.35±0.04	86.63±0.06	89.45±0.29	57.41±0.34
APL	88.15±0.09	81.51±0.17	86.07±0.09	51.66±0.20	77.34±0.23	32.66±0.26	78.83±2.11	60.00±1.71
ELR	89.42±0.11	87.52±0.10	89.87±0.10	89.36±0.16	88.96±0.09	88.48±0.14	89.78±0.13	90.05±0.17
MSLC	93.20±0.09	90.67±0.17	94.09±0.09	92.74±0.11	92.82±0.11	92.44±0.11	92.46±0.10	91.93±0.14
CDR	82.79±0.66	64.16±0.74	81.25±0.96	60.17±0.94	82.14±0.60	64.13±1.13	81.11±0.80	62.69±1.51
PES	92.48±0.06	89.17±0.13	92.14±0.03	88.77±0.09	92.40±0.05	88.70±0.08	91.57±0.07	89.27±0.12
MLC	89.46±0.12	86.53±0.24	90.33±0.15	90.30±0.14	90.31±0.15	90.29±0.13	89.74±0.13	89.51±0.14
CTRR	92.97±0.32	92.16±0.31	93.05±0.56	92.82±0.27	93.89±0.39	93.04±0.33	91.67±0.25	90.16±0.50
FPL	93.55±0.03	92.92±0.04	94.13±0.07	94.08±0.09	94.23±0.06	93.71±0.06	93.70±0.06	93.50±0.04
Meta-FPL	93.60±0.05	93.01±0.05	94.20±0.06	94.12±0.07	94.32±0.07	93.78±0.06	93.77±0.05	93.60±0.08

算法	Symmetric		Pairflip		Tridiagonal		Instance
算法	NR为20%	NR为40%	NR为20%	NR为40%	NR为20%	NR为40%	NR为20%	NR为40%
Standard	56.78±0.20	38.09±0.25	58.27±0.11	41.52±0.17	59.99±0.13	44.25±0.33	55.10±0.37	36.78±0.26
GCE	65.24±0.56	62.39± 0.38	66.21±0.09	57.62±0.12	64.79±0.11	58.35±0.13	61.87±0.39	47.66±0.69
APL	56.41±0.11	46.25±0.11	56.00±0.14	41.95±0.16	56.40±0.14	45.08±0.11	55.46±0.12	44.68±0.16
ELR	70.83±0.19	66.91±0.21	73.46±0.13	71.69±0.17	72.60±0.16	72.48±0.20	70.09±0.13	69.59±0.19
MSLC	71.47±0.20	67.50±0.14	72.62±0.13	71.79±0.24	72.33±0.08	71.71±0.12	70.82±0.11	69.04±0.15
CDR	58.45±0.27	52.56±0.34	63.62±0.26	45.82±0.25	65.39±0.31	48.89±0.35	61.70±0.56	43.00±0.85
PES	69.26±0.07	64.96±0.14	70.67±0.05	64.85±0.12	71.43±0.13	67.40±0.18	68.30±0.07	66.27±0.16
MLC	50.92±0.22	39.76±0.13	58.85±0.23	48.88±0.14	58.85±0.23	48.87±0.14	50.67±0.17	47.72±0.22
CTRR	70.09±0.45	65.32±0.20	72.88±0.34	66.75±0.28	73.12±0.42	64.46±0.17	70.15±0.23	64.53±0.17
FPL	71.55±0.08	67.84±0.04	73.43±0.09	72.35±0.09	73.26±0.03	72.62±0.07	71.49±0.05	70.34±0.05
Meta-FPL	71.58±0.06	67.86±0.07	73.47±0.07	72.41±0.09	73.24±0.05	72.68±0.08	71.44±0.06	70.38±0.05

算法	Symmetric		Pairflip		Tridiagonal		Instance
算法	NR为20%	NR为40%	NR为20%	NR为40%	NR为20%	NR为40%	NR为20%	NR为40%
Standard	56.78±0.20	38.09±0.25	58.27±0.11	41.52±0.17	59.99±0.13	44.25±0.33	55.10±0.37	36.78±0.26
GCE	65.24±0.56	62.39± 0.38	66.21±0.09	57.62±0.12	64.79±0.11	58.35±0.13	61.87±0.39	47.66±0.69
APL	56.41±0.11	46.25±0.11	56.00±0.14	41.95±0.16	56.40±0.14	45.08±0.11	55.46±0.12	44.68±0.16
ELR	70.83±0.19	66.91±0.21	73.46±0.13	71.69±0.17	72.60±0.16	72.48±0.20	70.09±0.13	69.59±0.19
MSLC	71.47±0.20	67.50±0.14	72.62±0.13	71.79±0.24	72.33±0.08	71.71±0.12	70.82±0.11	69.04±0.15
CDR	58.45±0.27	52.56±0.34	63.62±0.26	45.82±0.25	65.39±0.31	48.89±0.35	61.70±0.56	43.00±0.85
PES	69.26±0.07	64.96±0.14	70.67±0.05	64.85±0.12	71.43±0.13	67.40±0.18	68.30±0.07	66.27±0.16
MLC	50.92±0.22	39.76±0.13	58.85±0.23	48.88±0.14	58.85±0.23	48.87±0.14	50.67±0.17	47.72±0.22
CTRR	70.09±0.45	65.32±0.20	72.88±0.34	66.75±0.28	73.12±0.42	64.46±0.17	70.15±0.23	64.53±0.17
FPL	71.55±0.08	67.84±0.04	73.43±0.09	72.35±0.09	73.26±0.03	72.62±0.07	71.49±0.05	70.34±0.05
Meta-FPL	71.58±0.06	67.86±0.07	73.47±0.07	72.41±0.09	73.24±0.05	72.68±0.08	71.44±0.06	70.38±0.05