Lung cancer image classification method via bidirectional collaborative distillation and multi-instance learning

doi:10.11772/j.issn.1001-9081.2025080960

Journal of Computer Applications ›› 2026, Vol. 46 ›› Issue (9): 3005-3016.DOI: 10.11772/j.issn.1001-9081.2025080960

• Multimedia computing and computer simulation • Previous Articles

Lung cancer image classification method via bidirectional collaborative distillation and multi-instance learning

Jinchang LI¹, Wei JIA¹^,²(), Xuefen ZHAO¹^,²

^1.School of Information Engineering，Ningxia University，Yinchuan Ningxia 750021，China
^2.Ningxia Key Laboratory of Artificial Intelligence and Information Security for Channeling Computing Resources from the East to the West （Ningxia University），Yinchuan Ningxia 750021，China

Received:2025-08-22 Revised:2025-12-17 Accepted:2026-01-04 Online:2026-02-12 Published:2026-09-10
Contact: Wei JIA
About author:LI Jinchang， born in 2000， M. S. candidate. His research interests include medical image processing and analysis， computational pathology.
JIA Wei， born in 1980， Ph. D.， associate professor. His research interests include medical image processing and analysis， computational pathology.
ZHAO Xuefen， born in 1983， Ph. D.， associate professor. Her research interests include medical image processing and analysis.
Supported by:
Natural Science Foundation of Ningxia(2024AAC03063)

双向协同蒸馏多实例学习的肺癌图像分类方法

李锦昌¹, 贾伟¹^,²(), 赵雪芬¹^,²

^1.宁夏大学信息工程学院，银川 750021
^2.宁夏“东数西算”人工智能与信息安全重点实验室（宁夏大学），银川 750021

通讯作者: 贾伟
作者简介:李锦昌（2000—），男，江西宜春人，硕士研究生，主要研究方向：医学图像处理与分析、计算病理学
贾伟（1980—），男，宁夏银川人，副教授，博士，CCF会员，主要研究方向：医学图像处理与分析、计算病理学
赵雪芬（1983—），女，宁夏隆德人，副教授，博士，主要研究方向：医学图像处理与分析。
基金资助:
宁夏自然科学基金资助项目(2024AAC03063)

Abstract

Abstract:

To address the issue in the existing Whole Slide Image （WSI） classification for lung cancer， where Multi-Instance Learning （MIL） methods based on knowledge distillation architecture struggle to handle the mixture of hard and easy instances effectively， leading to insufficient model learning capability and imbalanced performance， a lung cancer whole slide image classification method based on Bidirectional Collaborative Distillation and Multi-Instance Learning （BCD-MIL） was proposed by incorporating the concept of multi-task learning. First， a Bidirectional Collaborative Distillation Framework （BCDF） comprising multiple student models and a teacher model was designed， so as to satisfy the differentiated learning requirements for easy and hard pathological instances in lung cancer histopathological images， and by distributing different tasks to multiple student models， the model's learning capacity was enhanced while solving performance imbalance problem. Concurrently， classification performance was improved through a virtuous cycle with the teacher model. Second， a Multi-Angle Instance Mining Parallel architecture （MA-IMP） was designed to match the heterogeneous differences in lung cancer pathological features across dimensions such as cell morphology and tissue texture， and by conducting instance mining from multiple perspectives， the mining bias caused by a single perspective was avoided. Finally， a Dynamic Stage-Aware Exponential Moving Average distillation （DSA-EMA） algorithm was proposed to optimize weight update of the teacher model， and improve training efficiency and model performance based on the stage characteristics of large-scale instance training of lung cancer histopathological images through adjusting distillation parameters in the training phase dynamically. Experimental results show that compared to the MIL framework with Masked Hard Instance Mining （MHIM-MIL） method， which is also based on knowledge distillation and instance mining， BCD-MIL achieves improvements of 1.92， 2.37， 3.43， and 1.75 percentage points in Area Under Curve （AUC）， accuracy， F1-Score， and recall， respectively， on The Cancer Genome Atlas （TCGA） dataset， and improvements of 1.11， 6.29， 6.94， and 12.85 percentage points in four metrics， respectively， on the Clinical Proteomic Tumor Analysis Consortium （CPTAC） dataset； validating the effectiveness of the proposed method. Furthermore， the lightweight distillation architecture reduces model parameter size and inference time， thereby enabling efficient deployment while ensuring performance gains， and providing a reliable basis for lung cancer WSI classification.

Key words: lung cancer, Whole Slide Image (WSI) classification, Multi-Instance Learning (MIL), knowledge distillation, instance mining

摘要：

针对现有的肺癌全切片图像（WSI）分类中，基于知识蒸馏架构的多实例学习（MIL）方法难以有效处理混杂的难分类和易分类实例，导致模型学习能力不足与性能不均衡的问题，结合多任务学习思想，提出一种双向协同蒸馏多实例学习的肺癌全切片图像分类方法（BCD-MIL）。首先，设计一种包括多学生模型与教师模型的双向协同蒸馏框架（BCDF），以满足肺癌病理图像中易分类与难分类病理实例的差异化学习需求，通过多个学生模型分担不同任务，增强模型的学习能力并解决性能不均衡的问题，同时，通过与教师模型的良性循环提升模型的分类性能；其次，设计一种多角度实例挖掘的并行结构（MA-IMP），以匹配肺癌病理特征在细胞形态和组织纹理等维度的异质性差异，通过从多个不同角度进行实例挖掘，避免单一角度造成的实例挖掘偏差；最后，提出一种动态阶段感知的指数移动平均蒸馏（DSA-EMA）算法，根据肺癌病理图像大规模实例训练的阶段特性，通过在训练阶段动态调整蒸馏参数，优化教师模型权重更新，提高训练效率与模型性能。实验结果表明，与同样基于知识蒸馏和实例挖掘的多实例学习方法的MHIM-MIL（MIL framework with Masked Hard Instance Mining）相比，BCD-MIL在TCGA（The Cancer Genome Atlas）和CPTAC（Clinical Proteomic Tumor Analysis Consortium）数据集上的曲线下面积（AUC）、准确率、F1-Score和召回率分别提升了1.92、2.37、3.43和1.75个百分点，以及1.11、6.29、6.94和12.85个百分点，验证了该方法的有效性；同时，轻量化的蒸馏架构降低了模型的参数规模与推理耗时，从而在保证性能提升的前提下实现了高效部署，为肺癌WSI分类提供了可靠依据。

关键词: 肺癌, 全切片图像分类, 多实例学习, 知识蒸馏, 实例挖掘

CLC Number:

TP391.4

Jinchang LI, Wei JIA, Xuefen ZHAO. Lung cancer image classification method via bidirectional collaborative distillation and multi-instance learning[J]. Journal of Computer Applications, 2026, 46(9): 3005-3016.

李锦昌, 贾伟, 赵雪芬. 双向协同蒸馏多实例学习的肺癌图像分类方法[J]. 《计算机应用》唯一官方网站, 2026, 46(9): 3005-3016.

Figures/Tables 14

Fig. 1 Overall structure of BCD-MIL

Fig. 2 Comparison of two knowledge distillation structures

Fig. 3 Structure of FCM

Fig. 4 Structure of MA-IMP

Tab. 1 Dataset sample partitioning sizes

数据集	训练集样本数	测试集样本数	验证集样本数
TCGA	687	263	103
CPTAC	1 448	554	216

Tab. 2 Experimental results of loss weight factor parameter on TCGA dataset

$η$	AUC/%	ACC/%	F1-Score/%	Precision/%	Recall/%
0.9	97.45	93.86	93.14	91.46	94.28
0.8	97.01	90.43	89.83	93.80	86.15
0.7	96.95	91.23	89.01	85.77	93.19
0.5	96.65	89.02	89.31	92.29	86.09
0.3	95.51	88.10	87.01	93.13	82.04
0.1	93.34	87.05	84.84	87.86	83.89

Tab. 2 Experimental results of loss weight factor parameter on TCGA dataset

$η$	AUC/%	ACC/%	F1-Score/%	Precision/%	Recall/%
0.9	97.45	93.86	93.14	91.46	94.28
0.8	97.01	90.43	89.83	93.80	86.15
0.7	96.95	91.23	89.01	85.77	93.19
0.5	96.65	89.02	89.31	92.29	86.09
0.3	95.51	88.10	87.01	93.13	82.04
0.1	93.34	87.05	84.84	87.86	83.89

Tab. 3 Experimental results of loss weight factor parameter on CPTAC dataset

$η$	AUC/%	ACC/%	F1-Score/%	Precision/%	Recall/%
0.9	97.34	92.46	92.78	93.87	92.10
0.8	97.16	90.16	91.41	92.98	90.26
0.7	96.78	89.48	90.08	92.54	88.00
0.5	95.54	85.14	86.53	91.69	81.46
0.3	94.71	86.45	84.86	92.67	79.02
0.1	91.09	81.44	80.01	87.31	75.44

Tab. 3 Experimental results of loss weight factor parameter on CPTAC dataset

$η$	AUC/%	ACC/%	F1-Score/%	Precision/%	Recall/%
0.9	97.34	92.46	92.78	93.87	92.10
0.8	97.16	90.16	91.41	92.98	90.26
0.7	96.78	89.48	90.08	92.54	88.00
0.5	95.54	85.14	86.53	91.69	81.46
0.3	94.71	86.45	84.86	92.67	79.02
0.1	91.09	81.44	80.01	87.31	75.44

Tab. 4 Experimental results of performance comparison on TCGA and CPTAC datasets

方法	TCGA					CPTAC
方法	AUC	ACC	F1-Score	Precision	Recall	AUC	ACC	F1-Score	Precision	Recall
AB-MIL^［17］	93.17	88.03	87.41	86.92	87.52	93.48	86.63	86.69	91.84	82.49
DSMIL^［19］	94.68	89.32	90.37	91.62	87.83	94.78	84.90	86.51	92.43	80.54
CLAM-SB^［18］	93.67	87.74	87.36	87.07	88.25	94.55	82.82	82.28	91.98	74.73
CLAM-MB^［18］	93.69	88.73	88.28	88.42	88.74	94.60	85.48	86.78	84.73	88.93
TransMIL^［21］	92.51	87.08	86.40	86.53	87.62	96.77	90.14	90.77	92.10	89.52
DTFD-MIL^［24］	93.83	88.23	87.71	87.12	88.16	96.71	89.20	89.52	93.97	85.53
MHIM-MIL^［30］	95.53	91.49	89.71	90.53	89.84	96.23	86.17	85.84	95.13	79.25
MambaMIL^［22］	96.84	88.14	88.23	89.76	86.85	96.97	89.03	88.72	87.36	89.10
TKA-MIL^［25］	96.78	89.13	88.84	90.37	87.36	96.64	89.76	89.46	92.65	86.67
BCD-MIL	97.45	93.86	93.14	94.95	91.59	97.34	92.46	92.78	93.87	92.10

Fig. 5 Comparison of confusion matrices in accuracy comparison experiments on TCGA and CPTAC datasets

Tab. 5 Lightweight comparison experimental results on TCGA and CPTAC datasets

方法	TCGA			CPTAC
方法	Para./10⁶	Time/ms	Mem./MB	Para./10⁶	Time/ms	Mem. /MB
AB-MIL^［17］	0.67	1.78	229.33	0.67	1.93	349.33
DSMIL^［19］	0.79	3.58	344.45	0.79	3.58	463.62
CLAM-SB^［18］	0.81	3.75	286.49	0.80	4.11	375.83
CLAM-MB^［18］	0.79	3.62	288.62	0.77	3.15	383.11
TransMIL^［21］	2.57	4.67	1 384.82	2.67	4.67	1 756.89
DTFD-MIL^［24］	0.98	1.68	213.26	0.96	2.60	329.33
MHIM-MIL^［30］	0.67	1.78	197.09	0.67	2.00	346.75
MambaMIL^［22］	1.48	2.47	687.98	1.53	2.56	958.64
TKA-MIL^［25］	0.81	1.89	267.43	0.82	2.13	390.82
BCD-MIL	0.67	1.77	195.10	0.67	1.89	344.45

Tab. 6 Ablation experimental results on two datasets

基准	BCDF	FCM	DSA-EMA	TCGA					CPTAC
基准	BCDF	FCM	DSA-EMA	AUC	ACC	F1-Score	Precision	Recall	AUC	ACC	F1-Score	Precision	Recall
􀳫				94.68	89.32	90.37	91.62	87.83	95.37	87.03	87.56	90.28	86.58
􀳫	􀳫			96.23	89.85	90.04	91.67	88.53	96.30	88.46	88.79	93.40	85.03
􀳫		􀳫		95.83	90.05	89.84	92.67	88.46	95.92	87.23	87.65	92.75	83.72
􀳫	􀳫	􀳫		97.02	91.96	92.83	92.90	92.15	96.68	88.77	89.51	90.62	88.70
􀳫	􀳫		􀳫	96.55	91.88	91.74	91.94	90.67	96.24	87.67	87.98	93.01	84.11
􀳫		􀳫	􀳫	96.43	91.43	90.82	92.80	89.13	97.06	90.42	91.46	90.31	92.76
􀳫	􀳫	􀳫	􀳫	97.45	93.86	93.14	94.95	91.95	97.34	92.46	92.78	93.87	92.10

Tab. 7 Experimental results of masking probability parameter

$r a v g$	TCGA					CPTAC
$r a v g$	AUC/%	ACC/%	F1-Score/%	Precision/%	Recall/%	AUC/%	ACC/%	F1-Score/%	Precision/%	Recall/%
0.442	96.33	92.64	92.45	93.52	91.21	96.58	91.26	91.57	92.33	90.66
0.484	97.45	93.86	93.14	94.95	91.59	97.34	92.46	92.78	93.87	92.10
0.539	96.18	92.52	92.63	93.46	91.51	96.86	91.33	91.65	92.42	90.79

Tab. 7 Experimental results of masking probability parameter

$r a v g$	TCGA					CPTAC
$r a v g$	AUC/%	ACC/%	F1-Score/%	Precision/%	Recall/%	AUC/%	ACC/%	F1-Score/%	Precision/%	Recall/%
0.442	96.33	92.64	92.45	93.52	91.21	96.58	91.26	91.57	92.33	90.66
0.484	97.45	93.86	93.14	94.95	91.59	97.34	92.46	92.78	93.87	92.10
0.539	96.18	92.52	92.63	93.46	91.51	96.86	91.33	91.65	92.42	90.79

Tab. 8 Comparison of consistency between MA-IMP mechanism and pathologist diagnosis

数据集	抽样数	Kappa系数	一致性
TCGA	150	0.80	高度一致
CPTAC	150	0.78	高度一致

Fig. 6 Visualization results of different models on two datasets

References 43

[1]	Siegel R L， Miller K D， Wagle N S， et al. Cancer statistics， 2023［J］. CA： A Cancer Journal for Clinicians， 2023， 73（1）： 17-48.
[2]	George J E， George P S， Krishna J K M， et al. Global trends in lung cancer incidence and mortality by age， gender and morphology and forecast： a bootstrap-based analysis ［J］. Lung Cancer， 2025， 205： No.108626.
[3]	Khuder S A. Effect of cigarette smoking on major histological types of lung cancer： a meta-analysis ［J］. Lung Cancer， 2001， 31（2/3）： 139-148.
[4]	Gao S， Li N， Wang S， et al. Lung cancer in People’s Republic of China ［J］. Journal of Thoracic Oncology， 2020， 15（10）： 1567-1576.
[5]	Han B， Zheng R， Zeng H， et al. Cancer incidence and mortality in China， 2022 ［J］. Journal of the National Cancer Center， 2024， 4（1）： 47-53.
[6]	Wang S， Zhao Y， Jha P， et al. Household air pollution and risk of incident lung cancer in urban China： a prospective cohort study ［J］. International Journal of Cancer， 2023， 153（9）： 1592-1601.
[7]	Zhuang J， Liu Z， Xiao R， et al. Association between smoking and environmental tobacco smoke with lung cancer risk： a case-control study in the Fujian Chinese population ［J］. Journal of Public Health， 2022， 30（8）： 2047-2057.
[8]	Evans A J， Brown R W， Bui M M， et al. Validating whole slide imaging systems for diagnostic purposes in pathology： guideline update from the College of American Pathologists in collaboration with the American Society for Clinical Pathology and the Association for Pathology Informatics ［J］. Archives of Pathology and Laboratory Medicine， 2022， 146（4）： 440-450.
[9]	Chen M， Liu X， Du J， et al. Differentiated regulation of immune-response related genes between LUAD and LUSC subtypes of lung cancers ［J］. Oncotarget， 2017， 8（1）： 133-144.
[10]	Sumon R I， Mazumdar M A I， Uddin S M I， et al. Exploring deep learning and machine learning techniques for histopathological image classification in lung cancer diagnosis ［C］// ICECET 2024. Piscataway： IEEE， 2024： 123-130.
[11]	Mohalder R D， Sarkar J P， Hossain K A， et al. A deep learning based approach to predict lung cancer from histopathological images ［C］// ICECIT 2021. Piscataway： IEEE， 2021： 1-4.
[12]	Maron O， Lozano-Pérez T. A framework for multiple-instance learning ［C］// NeurIPS 1997. Cambridge： MIT Press， 1997： 570-576.
[13]	Pantanowitz L， Sinard J H， Henricks W H， et al. Validating whole slide imaging for diagnostic purposes in pathology： guideline from the College of American Pathologists Pathology and Laboratory Quality Center ［J］. Archives of Pathology and Laboratory Medicine， 2013， 137（12）： 1710-1722.
[14]	Zarella M D， Bowman D， Aeffner F， et al. A practical guide to whole slide imaging： a white paper from the Digital Pathology Association ［J］. Archives of Pathology and Laboratory Medicine， 2019， 143（2）： 222-234.
[15]	Dietterich T， Lathrop R， Lozano-Pérez T. Solving the multiple instance problem with axis-parallel rectangles ［J］. Artificial Intelligence， 2001， 89（1/2）： 31-71.
[16]	Cheplygina V， de Bruijne M， Pluim J P W. Not-so-supervised： a survey of semi-supervised， multiple instance， and transfer learning in medical image analysis ［J］. Medical Image Analysis， 2019， 54： 280-296.
[17]	Ilse M， Tomczak J， Welling M. Attention-based deep multiple instance learning ［J］. Proceedings of Machine Learning Research， 2018， 80： 2127-2136.
[18]	Lu M Y， Williamson D F K， Chen T Y， et al. Data-efficient and weakly supervised computational pathology on whole-slide images［J］. Nature Biomedical Engineering， 2021， 5（6）： 555-570.
[19]	Li B， Li Y， Eliceiri K W， et al. Dual-stream multiple instance learning network for whole slide image classification with self-supervised contrastive learning ［C］// CVPR 2021. Piscataway： IEEE， 2021： 14313-14323.
[20]	Vaswani A， Shazeer N， Parmar N， et al. Attention is all you need［C］// NeurIPS 2017. Red Hook： Curran Associates Inc.， 2017： 6000-6010.
[21]	Shao Z， Bian H， Chen Y， et al. TransMIL： Transformer based correlated multiple instance learning for whole slide image classification ［C］// NeurIPS 2021. Red Hook： Curran Associates Inc.， 2021： 1236-1247.
[22]	Yang S， Wang Y， Chen H. MambaMIL： enhancing long sequence modeling with sequence reordering in computational pathology ［PP/OL］. arXiv （2024-03-11）［2025-05-15］..
[23]	Gu A， Dao T. Mamba： linear-time sequence modeling with selective state spaces ［PP/OL］. V2. arXiv （2024-05-31）［2025-05-14］..
[24]	Zhang H， Meng Y， Zhao Y， et al. DTFD-MIL： double-tier feature distillation multiple instance learning for histopathology whole slide image classification ［C］// CVPR 2022. Piscataway： IEEE， 2022： 18780-18790.
[25]	Yu S， Chen X， Cao F， et al. TKA-MIL： top-k attention multiple instance learning for whole slide image classification and instance probability derivation ［C］// ICASSP 2025. Piscataway： IEEE， 2025： 1-5.
[26]	Hinton G， Vinyals O， Dean J， et al. Distilling the knowledge in a neural network［PP/OL］. arXiv （2015-03-09）［2025-05-13］..
[27]	Wu J， Fang L Y， Yue J. TAKD： target-aware knowledge distillation for remote sensing scene classification ［J］. IEEE Transactions on Circuits and Systems for Video Technology， 2024， 34（9）： 8188-8200.
[28]	Tisha F T， Alam M， Mishu S Z， et al. Deep reinforcement learning with mutual information for selecting bands in hyperspectral image classification ［C］// PEEIACON 2024. Piscataway： IEEE， 2024： 804-809.
[29]	Shen G， Shen Y， Riaz M N. Training deep face recognition for efficient inference by distillation and mutual learning ［C］// PIC 2018. Piscataway： IEEE， 2018： 38-43.
[30]	Tang W， Huang S， Zhang X， et al. Multiple instance learning framework with masked hard instance mining for whole slide image classification ［C］// ICCV 2023. Piscataway： IEEE， 2023： 4055-4064.
[31]	Keshvarikhojasteh H， Pluim J P W， Veta M. Multi-head attention-based deep multiple instance learning ［PP/OL］. arXiv （2024-04-08）［2025-05-15］..
[32]	Liu D， Li C， Hu X， et al. Dual-attention multiple instance learning framework for pathology whole-slide image classification［J］. Electronics， 2024， 13（22）： No.4445.
[33]	Lin T， Yu Z， Hu H， et al. Interventional bag multi-instance learning on whole-slide pathological images ［C］// CVPR 2023. Piscataway： IEEE， 2023： 19830-19839.
[34]	Do K， Le H， Nguyen D， et al. Momentum adversarial distillation： handling large distribution shifts in data-free knowledge distillation［C］// NeurIPS 2022. Red Hook： Curran Associates Inc.， 2022： 10055-10067.
[35]	Tian J， Lyu F. Parameter-selective continual test-time adaptation［C］// ACCV 2024， LNCS 15479. Singapore： Springer， 2025： 315-331.
[36]	吕君环，许柯，王东. 基于指数移动平均知识蒸馏的神经网络低比特量化方法［J］. 模式识别与人工智能， 2021， 34（12）： 1143-1151.
	Junhuan Lyu， Xu Ke， Wang Dong. Low-bit quantization of neural network based on exponential moving average knowledge distillation ［J］. Pattern Recognition and Artificial Intelligence， 2021， 34 （12）： 1143-1151.
[37]	Deng J， Dong W， Socher R， et al. ImageNet： a large-scale hierarchical image database ［C］// CVPR 2009. Piscataway： IEEE， 2009： 248-255.
[38]	He K， Zhang X， Ren S， et al. Deep residual learning for image recognition ［C］// CVPR 2016. Piscataway： IEEE， 2016： 770-778.
[39]	Kingma D P， Ba J L. Adam： a method for stochastic optimization［PP/OL］. V9. arXiv （2017-01-30）［2025-05-24］..
[40]	黄震华，杨顺志，林威，等. 知识蒸馏研究综述［J］. 计算机学报， 2022， 45（3）： 624-653.
	Huang Zhenhua， Yang Shunzhi， Lin Wei， et al. Knowledge distillation： a survey ［J］. Chinese Journal of Computers， 2022， 45（3）： 624-653.
[41]	邵仁荣，刘宇昂，张伟，等. 深度学习中知识蒸馏研究综述［J］. 计算机学报， 2022， 45（8）： 1638-1673.
	Shao Renrong， Liu Yuang， Zhang Wei， et al. A survey of knowledge distillation in deep learning ［J］. Chinese Journal of Computers， 2022， 45（8）： 1638-1673.
[42]	Xie Y， Liu Z， Zhao J， et al. PHIM-MIL： multiple instance learning with prototype similarity-guided feature fusion and hard instance mining for whole slide image classification ［J］. Information Fusion， 2025， 117： No.102847.
[43]	Martín Andrés A， Álvarez Hernández M. Estimators of various kappa coefficients based on the unbiased estimator of the expected index of agreements ［J］. Advances in Data Analysis and Classification， 2025， 19（1）： 177-207.

Lung cancer image classification method via bidirectional collaborative distillation and multi-instance learning

双向协同蒸馏多实例学习的肺癌图像分类方法

RichHTML

PDF

Knowledge

Abstract

Cite this article

share this article

Figures/Tables 14

References 43

Related Articles 15

Recommended Articles

Metrics

[1]	Yuelan ZHANG, Jing SU, Hangyu ZHAO, Baili YANG. Multi-view knowledge-aware and interactive distillation recommendation algorithm [J]. Journal of Computer Applications, 2025, 45(7): 2211-2220.
[2]	Sijie NIU, Yuliang LIU. Auxiliary diagnostic method for retinopathy based on dual-branch structure with knowledge distillation [J]. Journal of Computer Applications, 2025, 45(5): 1410-1414.
[3]	Sheng YANG, Yan LI. Contrastive knowledge distillation method for object detection [J]. Journal of Computer Applications, 2025, 45(2): 354-361.
[4]	Hu WANG, Xiaofeng WANG, Ke LI, Yunjie MA. Federated class-incremental learning method of label semantic embedding with multi-head self-attention [J]. Journal of Computer Applications, 2025, 45(10): 3083-3090.
[5]	Jieru JIA, Jianchao YANG, Shuorui ZHANG, Tao YAN, Bin CHEN. Unsupervised person re-identification based on self-distilled vision Transformer [J]. Journal of Computer Applications, 2024, 44(9): 2893-2902.
[6]	Yubo ZHAO, Liping ZHANG, Sheng YAN, Min HOU, Mao GAO. Relation extraction between discipline knowledge entities based on improved piecewise convolutional neural network and knowledge distillation [J]. Journal of Computer Applications, 2024, 44(8): 2421-2429.
[7]	Xue LI, Guangle YAO, Honghui WANG, Jun LI, Haoran ZHOU, Shaoze YE. Remote sensing image classification based on sample incremental learning [J]. Journal of Computer Applications, 2024, 44(3): 732-736.
[8]	Yuxin HUANG, Yiwang HUANG, Hui HUANG. Meta label correction method based on shallow network predictions [J]. Journal of Computer Applications, 2024, 44(11): 3364-3370.
[9]	Xujian ZHAO, Hanglin LI. Deep neural network compression algorithm based on hybrid mechanism [J]. Journal of Computer Applications, 2023, 43(9): 2686-2691.
[10]	Zhangjian JI, Ming ZHANG, Zilong WANG. High-precision object detection algorithm based on improved VarifocalNet [J]. Journal of Computer Applications, 2023, 43(7): 2147-2154.
[11]	Huaiqing HE, Jianqing YAN, Kanghua HUI. Lightweight face recognition method based on deep residual network [J]. Journal of Computer Applications, 2022, 42(7): 2030-2036.
[12]	Wei REN, Hexiang BAI. Multi-label image classification method based on global and local label relationship [J]. Journal of Computer Applications, 2022, 42(5): 1383-1390.
[13]	Junhua GU, Shuai FAN, Ningning LI, Suqi ZHANG. Long- and short-term recommendation model and updating method based on knowledge graph preference attention network [J]. Journal of Computer Applications, 2022, 42(4): 1079-1086.
[14]	ZHANG Cheng, WAN Yuan, QIANG Haopeng. Deep unsupervised discrete cross-modal hashing based on knowledge distillation [J]. Journal of Computer Applications, 2021, 41(9): 2523-2531.
[15]	HUANG Jishuang, ZHANG Hua, LI Yonglong, ZHAO Hao, WANG Haoran, FENG Chuncheng. Hydraulic tunnel defect recognition method based on dynamic feature distillation [J]. Journal of Computer Applications, 2021, 41(8): 2358-2365.