计算机科学

计算机科学 ›› 2024, Vol. 51 ›› Issue (8): 152-159.doi: 10.11896/jsjkx.230500066

• 计算机图形学&多媒体 • 上一篇    下一篇

基于伪标签依赖增强与噪声干扰消减的小样本图像分类

唐芮琪, 肖婷, 迟子秋, 王喆   

  1. 华东理工大学信息科学与工程学院 上海 200237
  • 收稿日期:2023-05-10 修回日期:2023-09-23 出版日期:2024-08-15 发布日期:2024-08-13
  • 通讯作者: 王喆(wangzhe@ecust.edu.cn)
  • 作者简介:(18516660531@163.com)
  • 基金资助:
    上海市科技计划项目(21511100800&20511100600);国家自然科学基金(62076094)

Few-shot Image Classification Based on Pseudo-label Dependence Enhancement and NoiseInterferenceReduction

TANG Ruiqi, XIAO Ting, CHI Ziqiu, WANG Zhe   

  1. School of Information Science and Engineering,East China University of Science and Technology,Shanghai 200237,China
  • Received:2023-05-10 Revised:2023-09-23 Online:2024-08-15 Published:2024-08-13
  • About author:TANG Ruiqi,born in 1999,postgra-duate.Her main research interests include few-shot learning and image processing.
    WANG Zhe,born in 1981,Ph.D,professor,is a member of CCF(No.16666M).His main research interests include pattern recognition and image processing.
  • Supported by:
    Shanghai Science and Technology Program(21511100800&20511100600) and National Natural Science Foundation of China(62076094).

PDF (PC)

365

摘要： 深度学习在图像分类领域的成功很大程度上依赖于大规模数据,然而在许多应用场景中,收集足够的数据用于模型的训练是比较困难的。因此,旨在利用有限的数据获得高性能模型的小样本学习成为热点研究方向。在小样本图像分类领域,使用无标签数据来扩充训练数据集是一种常用的方法,但该方法面临两个亟待解决的难题:如何获取无标签数据的伪标签以及如何减轻噪声标签累积的负面影响?首先,为获得高质量的伪标签,需要解决由源域和目标域的分布偏移导致的噪声标签问题,因而提出基于希尔伯特-施密特独立准则(Hilbert-Schmidt Independent Criterion,HSIC)的依赖增强方法,通过最大化图像特征表示与标签之间的相关性,从而提高伪标签的预测可靠度。其次,为克服标签预测误差随着时间推移不断累积的问题,提出噪声标签干扰消减(Noise Label Interference Reduction,NLIR)方法,确保具有正确标签的样本的梯度始终主导着训练动态,从而将模型引向最优解。所提方法在小样本图像分类基准数据集mini-ImageNet和tiered-ImageNet上进行了评估,实验结果表明,该方法能够很好地利用无标签数据提升分类精度,具有良好的分类性能。

关键词: 深度学习, 图像分类, 小样本学习, 伪标签, 噪声标签, 希尔伯特-施密特独立准则

Abstract: The success of deep learning in image classification relies heavily on large-scale data.However,in many application scenarios,it is difficult to collect enough data for model training.Therefore,few-shot learning aimed at obtaining high-performance models with limited data becomes a hot research direction.In the field of few-shot image classification,using unlabeled data to augment the training datasets is a common method,but it faces two urgent problems:how to obtain pseudo-labels of unlabeled data and how to mitigate the negative impact of accumulated noise labels？ Firstly,in order to obtain high-quality pseudo-labels,it is necessary to solve the problem of noise labels caused by the distribution shift of the source domain and the target domain.A dependence enhancement method based on Hilbert-Schmidt independent criterion is proposed to improve the prediction reliability of pseudo-labels by maximizing the correlation between image feature representation and labels.Secondly,in order to overcome the problem of label prediction error that accumulates over time,a noise label interference reduction method is proposed to ensure that the gradient of samples with correct labels always dominates the training dynamics,so as to lead the model to the optimal solution.The above methods are evaluated on the benchmark datasets for few-shot image classification,namely mini-ImageNet and tiered-ImageNet.The results demonstrate that the proposed approach effectively utilizes unlabeled data to improve classification accuracy and achieves impressive classification performance.

Key words: Deep learning, Image classification, Few-shot learning, Pseudo-labels, Noise-labels, Hilbert-Schmidt independent criterion

中图分类号: 

  • TP391
[1]ZHAO K L,JIN X L,WANG Y Z.A Review of Few-shotLearning[J].Journal of Software,2021,32(2):349-369.
[2]CHEN W Y,LIU Y C,KIRA Z,et al.A Closer Look at Few-shot Classification[J].arXiv:1904.04232,2019.
[3]RAINA R,BATTLE A,LEE H,et al.Self-taught Learning:Transfer Learning from Unlabeled Data[C]//Proceedings of the 24th International Conference on Machine Learning.Corvallis,USA:ACM,2007:759-766.
[4]LIANG J,HU D,FENG J.Do We Really Need to Access the Source Data? Source Hypothesis Transfer for Unsupervised Domain Adaptation[C]//Proceedings of the 37th International Conference on Machine Learning.Online:PMLR,2020:6028-6039.
[5]FINN C,ABBEEL P,LEVINE S.Model-agnostic Meta-learning for Fast Adaptation of Deep Networks[C]//Proceedings of the 34th International Conference on Machine Learning.Sydney,Australia:PMLR,2017:1126-1135.
[6]VINYALS O,BLUNDELL C,LILLICRAP T,et al.MatchingNetworks for One Shot Learning[C]//Proceedings of the 30th International Conference on Neural Information Processing Systems.Barcelona,Spain:MIT Press,2016:3630-3638.
[7]SNELL J,SWERSKY K,ZEMEL R.Prototypical Networks for Few-shot Learning[C]//Proceedings of the 31th International Conference on Neural Information Processing Systems.Long Beach,USA:ACM,2017:4080-4090.
[8]ORESHKIN B,RODRÍGUEZ LÓPEZ P,LACOSTE A.Tadam:Task Dependent Adaptive Metric for Improved Few-shot Lear-ning[C]//Proceedings of the 32th International Conference on Neural Information Processing Systems.Montréal,Canada:MIT Press,2018:721-731.
[9]SUNG F,YANG Y,ZHANG L,et al.Learning to Compare:Relation Network for Few-shot Learning[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition.Salt Lake City,USA:IEEE Press,2018:1199-1208.
[10]ZHANG R,CHE T,GHAHRAMAN Z,et al.Metagan:An Adversarial Approach to Few-shot Learning[C]//Proceedings of the 32th International Conference on Neural Information Processing Systems.Montréal,Canada:MIT Press,2018:2365-2374.
[11]LI K,ZHANG Y L,LI K,et al.Adversarial Feature Hallucination Networks for Few-shot Learning[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition.New York,USA:IEEE Press,2020:13470-13479.
[12]LIU Y,LEE J,PARK M,et al.Learning to Propagate Labels:Transductive Propagation Network for Few-shot Learning[J].arXiv:1805.10002,2018.
[13]RODRÍGUEZ P,LARADJI I,DROUIN A,et al.EmbeddingPropagation:Smoother Manifold for Few-shot Classification[C]//European Conference on Computer Vision.Graz,Germany:Springer,2020:121-138.
[14]KYE S M,LEE H B,KIM H,et al.Meta-learned Confidence for Few-shot Learning[J].arXiv:2002.12017,2020.
[15]HOU R,CHANG H,MA B,et al.Cross Attention Network for Few-shot Classification[C]//Proceedings of the 33th International Conference on Neural Information Processing Systems.Vancouver Canada:MIT Press,2019:4005-4016.
[16]HU S X,MORENO P G,XIAO Y,et al.Empirical Bayes Transductive Meta-learning with Synthetic Gradients[C]//Procee-dings of the 8th International Conference on Learning Representations.Addis Ababa,Ethiopia:OpenReview.net,2020.
[17]DHILLON G S,CHAUDHARI P,RAVICHANDRAN A,et al.A Baseline for Few-shot Image Classification[C]//Proceedings of the 8th International Conference on Learning Representations.Addis Ababa,Ethiopia:OpenReview.net,2020.
[18]LI X,SUN Q,LIU Y,et al.Learning to Self-Train for Semi-Supervised Few-Shot Classification[C]//Proceedings of the 33th International Conference on Neural Information Processing Systems.Online:MIT Press,2019:10276-10286.
[19]REN M,TRIANTAFILLOU E,RAVI S,et al.Meta-learningfor Semi-supervised Few-shot Classification[J].arXiv:1803.00676,2018.
[20]WANG Y,XU C,LIU C,et al.Instance Credibility Inference for Few-shot Learning[C]//Proceedings of the IEEE/CVF Confe-rence on Computer Vision and Pattern Recognition.New York,USA:IEEE Press,2020:12836-12845.
[21]GREENFELD D,SHALIT U.Robust Learning with the Hilbert-schmidt Independence Criterion[C]//Proceedings of the 37th International Conference on Machine Learning.Online:PMLR,2020:3759-3768.
[22]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.Vancouver,Canada:MIT Press,2020:20331-20342.
[23]BOUVEYRON C,GIRARD S,SCHMID C.High-dimensionalData Clustering[J].Computational Statistics & Data Analysis,2007,52(1):502-519.
[24]FRALEY C,RAFTERY A E.Model-based Clustering,Discriminant Analysis,and Density Estimation[J].Journal of the American Statistical Association,2002,97(458):611-631.
[25]MCNICHOLAS P D,MURPHY T B.Model-based Clustering of Microarray Expression Data via Latent Gaussian Mixture Mo-dels[J].Bioinformatics,2010,26(21):2705-2712.
[26]LAINE S,AILA T.Temporal Ensembling for Semi-supervised Learning[J].arXiv:1610.02242,2016.
[27]RUSSAKOVSKY O,DENG J,SU H,et al.Imagenet LargeScale Visual Recognition Challenge[J].International Journal of Computer Vision,2015,115(3):211-252.
[28]MA R,FANG P,DRUMMOND T,et al.Adaptive PoincaréPoint to Set Distance for Few-shot Classification[C]//Procee-dings of the 36th AAAI Conference on Artificial Intelligence.Los Angeles,USA:AAAI,2022:1926-1934.
[29]ZHOU F,ZHANG L,WEI W.Meta-generating Deep Attentive Metric for Few-shot Classification[J].IEEE Transactions on Circuits and Systems for Video Technology,2022,32(10):6863-6873.
[30]YANG L,LI L,ZHANG Z,et al.DPGN:Distribution Propagation Graph Network for Few-shot Learning[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition.New York,USA:IEEE Press,2020:13390-13399.
[31]SIMON C,KONIUSZ P,NOCK R,et al.Adaptive Subspaces for Few-shot Learning[C]//Proceedings of the IEEE/CVF Confe-rence on Computer Vision and Pattern Recognition.Seattle,USA:IEEE Press,2020:4136-4145.
[32]ZHU H,KONIUSZ P.EASE:Unsupervised Discriminant Subspace Learning for Transductive Few-shot Learning[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition.New York,USA:IEEE Press,2022:9078-9088.
[33]YU Z,CHEN L,CHENG Z,et al.Transmatch:A Transfer-learning Scheme for Semi-supervised Few-shot Learning[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition.New York,USA:IEEE Press,2020:12856-12864.
[34]KOBAK D,LINDERMAN G C.Initialization is Critical for Preserving Global Data Structure in Both t-SNE and UMAP[J].Nature Biotechnology,2021,39(2):156-157.
Viewed
Full text


Abstract

Cited
  Shared   
  Discussed   
No Suggested Reading articles found!
渝ICP备16013121号-4
地址：重庆市渝北区洪湖西路18号    邮编：401121  
电话：023-63500828（编辑部） 023-67039612（会议/专题） 023-67039623（财务/发票）
E-mail：jsjkx12@163.com
本系统由北京玛格泰克科技发展有限公司设计开发