类内-类间通道注意力少样本分类

杨利平; 张天洋; 王宇阳; 辜小花

doi:10.37188/OPE.20233121.3145

您当前的位置：

首页 >

文章列表页 >

类内-类间通道注意力少样本分类

信息科学 | 更新时间：2023-11-16

- 类内-类间通道注意力少样本分类
- Intra-inter channel attention for few-shot classification
- 光学精密工程 2023年31卷第21期页码：3145-3155
- 作者机构：
  
  1.重庆大学光电技术及系统教育部重点实验室，重庆 400044
  2.重庆科技学院电气工程学院，重庆 401331
- 作者简介：
  
  [ "杨利平（1981-），男，内蒙古鄂尔多斯人，博士，副教授，硕士生导师，主要研究方向为机器学习、模式识别和信号处理。E-mail：yanglp@cqu.edu.cn" ]
  张天洋（1998-），男，重庆綦江人，硕士研究生，2020年于重庆大学获得学士学位，主要从事少样本学习在图像分类和声音事件检测中应用的研究。E-mail：zhangty@cqu.edu.cn
- 基金信息：
  
  国家重点研发计划课题(2022YFF0708001);重庆市自然科学基金面上项目(csts2021jcyj-msxmX0478)
- DOI：10.37188/OPE.20233121.3145
  中图分类号：
扫描看全文
杨利平,张天洋,王宇阳等.类内-类间通道注意力少样本分类[J].光学精密工程,2023,31(21):3145-3155.

YANG Liping,ZHANG Tianyang,WANG Yuyang,et al.Intra-inter channel attention for few-shot classification[J].Optics and Precision Engineering,2023,31(21):3145-3155.
杨利平,张天洋,王宇阳等.类内-类间通道注意力少样本分类[J].光学精密工程,2023,31(21):3145-3155. DOI： 10.37188/OPE.20233121.3145.

YANG Liping,ZHANG Tianyang,WANG Yuyang,et al.Intra-inter channel attention for few-shot classification[J].Optics and Precision Engineering,2023,31(21):3145-3155. DOI： 10.37188/OPE.20233121.3145.

摘要

针对元学习少样本分类样本特征鉴别能力不足的问题，提出了一种类内-类间通道注意力少样本分类方法（Intra-inter Channel Attention Few-shot Classification， ICAFSC）。ICAFSC在原型网络基础上设计了一个类内-类间通道注意力模块，该模块通过类内-类间距离度量计算通道权重实现特征加权，提高特征对类别的鉴别能力。为了克服直接在元训练阶段学习类内-类间通道注意力模块容易出现过拟合或欠拟合现象的问题，ICAFSC在原型网络的元训练之前增加一个预训练阶段。该阶段设计具有大量标记样本的分类任务，并利用这些任务充分训练类内-类间通道注意力模块，促使该模块达到较优的状态。在原型网络的元训练和元测试阶段，ICAFSC冻结类内-类间通道注意力模块的参数，分别实现少样本分类经验的学习与迁移。在MiniImagenet数据集上分别开展了1-shot和5-shot的少样本分类实验。实验结果表明：本文提出的类内-类间通道注意力少样本分类方法与原型网络相比，在1-shot和5-shot条件下分类准确率分别提高了1.93%和1.15%。

Abstract

As only one or a few training samples are used for few-shot classification tasks， the features extracted via a prototypical network cannot guarantee much discriminative power. Accordingly， this paper proposes an intra-inter channel attention few-shot classification （ICAFSC） method. This method uses an intra–inter channel attention module （ICAM） to calculate channel weights based on an intra-inter distance metric. The module is integrated into the prototypical network to make the embedded features more discriminative. To overcome the problems of overfitting or underfitting when directly learning the ICAM in the few-shot classification's meta-training stage， ICAFSC adds a pre-training stage before the meta-training of the prototypical network. We design adequate classification tasks with a large number of labeled samples to learn optimal parameters of the ICAM in the pre-training stage. Subsequently， in the meta-training and meta-testing stages of the prototypical network， ICAFSC first freezes the parameters of the ICAM to guarantee a stable channel attention relationship. It then achieves few-shot classification experience learning and transfer via meta-training and meta-testing. We conduct 1-shot and 5-shot few-shot classification experiments on the MiniImagenet dataset. The experimental results indicate that， compared to the prototypical network， the proposed ICAFSC method shows improvements of 1.93% and 1.15% for the 1-shot and 5-shot scenarios， respectively.

关键词

深度学习少样本分类元学习原型网络通道注意力

Keywords

deep learningfew-shot classificationmeta-learningprototypical networkchannel attention

references

JIANG K， ZHU L， SUN Q D. Joint dual-structural constrained and non-negative analysis representation learning for pattern classification［J］. Applied Artificial Intelligence， 2023， 37（1）： 2180821. doi: 10.1080/08839514.2023.2180821http://dx.doi.org/10.1080/08839514.2023.2180821

AGARWAL N， SONDHI A， CHOPRA K， et al. Transfer Learning： Survey and Classification［M］.Smart Innovations in Communication and Computational Sciences. Singapore： Springer Singapore， 2020： 145-155. doi: 10.1007/978-981-15-5345-5_13http://dx.doi.org/10.1007/978-981-15-5345-5_13

LI X， SUN Z， XUE J H， et al. A concise review of recent few-shot meta-learning methods［J］. Neurocomputing， 2021， 456： 463-468. doi: 10.1016/j.neucom.2020.05.114http://dx.doi.org/10.1016/j.neucom.2020.05.114

SONG Y， WANG T， MONDAL S K， et al. A Comprehensive Survey of Few-shot Learning： Evolution， Applications， Challenges， and Opportunities［EB/OL］. https：//arxiv.org/abs/2205.06743https://arxiv.org/abs/2205.06743， 2022-3-24. doi: 10.1145/3582688http://dx.doi.org/10.1145/3582688

LU J， GONG P， YE J， et al. Learning from Very Few Samples： a Survey［EB/OL］. https：//arxiv.org/abs/2007.15484https://arxiv.org/abs/2007.15484， 2022-7-30. doi: 10.1016/j.patcog.2023.109480http://dx.doi.org/10.1016/j.patcog.2023.109480

HOSPEDALES T， ANTONIOU A， MICAELLI P， et al. Meta-learning in neural networks： a survey［J］. IEEE Transactions on Pattern Analysis and Machine Intelligence， 2022， 44（9）： 5149-5169.

HUISMAN M， RIJN J N， PLAAT A. A survey of deep meta-learning［J］.Artificial Intelligence Review， 2021， 54（6）： 4483-4541. doi: 10.1007/s10462-021-10004-4http://dx.doi.org/10.1007/s10462-021-10004-4

SNELL J， SWERSKY K， ZEMEL R. Prototypical networks for few-shot learning［C］.Proceedings of the 31st International Conference on Neural Information Processing Systems. New York： ACM， 2017： 4080-4090.

YAN S P， ZHANG S Y， HE X M. A dual attention network with semantic embedding for few-shot learning［J］. Proceedings of the AAAI Conference on Artificial Intelligence， 2019， 33（1）： 9079-9086. doi: 10.1609/aaai.v33i01.33019079http://dx.doi.org/10.1609/aaai.v33i01.33019079

LEE S， MOON W， HEO J P. Task discrepancy maximization for fine-grained few-shot classification［C］.2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition （CVPR）.18-24，2022， New Orleans， LA， USA. IEEE， 2022： 5321-5330. doi: 10.1109/cvpr52688.2022.00526http://dx.doi.org/10.1109/cvpr52688.2022.00526

GIDARIS S， BURSUC A， KOMODAKIS N， et al. Boosting few-shot visual learning with self-supervision［C］.2019 IEEE/CVF International Conference on Computer Vision （ICCV）. October 27 - November 2， 2019， Seoul， Korea （South）. IEEE， 2020： 8058-8067. doi: 10.1109/iccv.2019.00815http://dx.doi.org/10.1109/iccv.2019.00815

HU J， SHEN L， SUN G. Squeeze-and-excitation networks［C］.2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition. 18-23，2018， Salt Lake City， UT， USA. IEEE， 2018： 7132-7141. doi: 10.1109/cvpr.2018.00745http://dx.doi.org/10.1109/cvpr.2018.00745

GUO M H， XU T X， LIU J J， et al. Attention mechanisms in computer vision： a survey［J］.Computational Visual Media， 2022， 8（3）： 331-368. doi: 10.1007/s41095-022-0271-yhttp://dx.doi.org/10.1007/s41095-022-0271-y

VINYALS O， BLUNDELL C， LILLICRAP T， et al. Matching networks for one shot learning［C］.Proceedings of the 30th International Conference on Neural Information Processing Systems. New York： ACM， 2016： 3637-3645.

RAVI S， LAROCHELLE H. Optimization as a model for few-shot learning［C］. Proceedings of the International Conference on Learning Representations， 2017.

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 - Volume 70. August 6 - 11， 2017， Sydney， NSW， Australia. New York： ACM， 2017： 1126-1135. doi: 10.1109/icra.2016.7487173http://dx.doi.org/10.1109/icra.2016.7487173

LIU X， ZHOU F， LIU J， et al. Meta-Learning based prototype-relation network for few-shot classification［J］. Neurocomputing， 2020， 383： 224-234. doi: 10.1016/j.neucom.2019.12.034http://dx.doi.org/10.1016/j.neucom.2019.12.034

OH J， YOO H， KIM C， et al. BOIL： Towards Representation Change for Few-shot Learning［EB/OL］. 2020： arXiv： 2008.08882. https：//arxiv.org/abs/2008.08882.pdfhttps://arxiv.org/abs/2008.08882.pdf.

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. doi: 10.1109/tcsvt.2022.3173687http://dx.doi.org/10.1109/tcsvt.2022.3173687

AN Y X， XUE H， ZHAO X Y， et al. Conditional self-supervised learning for few-shot classification［C］.Proceedings of the Thirtieth International Joint Conference on Artificial Intelligence. August 19-27， 2021. Montreal， Canada. California： International Joint Conferences on Artificial Intelligence Organization， 2021： 2140-2146.

BERTINETTO L， HENRIQUES J F， TORR P H S， et al. Meta-learning with differentiable closed-form solvers［C］. Proceedings of the International Conference on Learning Representations， 2019.

DING N， CHEN Y， CUI G， et al. Few-shot Classification with Hypersphere Modeling of Prototypes［EB/OL］. ［2022-11-10］. https：//arxiv.53yu.com/abs/2211.05319https://arxiv.53yu.com/abs/2211.05319. doi: 10.18653/v1/2023.findings-acl.57http://dx.doi.org/10.18653/v1/2023.findings-acl.57

SNELL J， RICHARD Z. Bayesian few-shot Classification with One-vs-each Pólya-gamma Augmented Gaussian Processes［EB/OL］. ［2021-1-21］. https：//arxiv.org/abs/2007.10417https://arxiv.org/abs/2007.10417.

SENDERA M， PRZEWIĘŹLIKOWSKI M， KARANOWSKI K， et al. HyperShot： few-shot learning by kernel HyperNetworks［C］.2023 IEEE/CVF Winter Conference on Applications of Computer Vision （WACV）.2-7，2023， Waikoloa， HI， USA. IEEE， 2023： 2468-2477. doi: 10.1109/wacv56688.2023.00250http://dx.doi.org/10.1109/wacv56688.2023.00250

ALLEN K， SHELHAMER E， SHIN H， et al. Infinite mixture prototypes for few-shot learning［C］. Proceedings of the International Conference on Machine Learning， 2019： 232-241.

浏览量

下载量

CSCD

文章被引用时，请邮件提醒。

提交

工具集

关联资源

融合卷积块注意力模块和Siamese神经网络的人脸识别算法

基于无监督学习和注意力机制的水下偏振图像融合

基于LL-GG-LG Net的CT和PET医学图像融合

零部件光学影像精准定位的轻量化深度学习网络

基于深度学习的多视图立体重建方法综述