基于多种数据集的自监督学习的少样本图像分类

左 宗侑, 李 丹
四川大学锦城学院

摘要

少样本图片分类的目的是用有限的标记样本对没有标记的类别进行分类。由于是对每个任务的样本数量限制，元学习的初始嵌入网络是元学习的重要组成部分并且在实际应用中会对元学习的性能产生很大的影响。因此，许多预先训练的方法被人们提了出来，其中大部分分类都是在监督的方式下进行训练的。在本文中，我们提出训练一个更为广义的嵌入网络与通过从数据本身学习来为下游任务提供表示的自监督学习(SSL)。我们通过在1-shot和5-shot的参数下对多种数据集训练的结果进行比较来评估我们的工作。

关键词

自监督学习；少样本学习；嵌入网络；图像分类

参考

[1]OriolVinyals,CharlesBlundell,TimothyLillicrap,DaanWierstra, et al., “Matching networks for one shot learning,” in NeurIPS, 2016, pp. 3630–3638.

[2] Sachin Ravi and Hugo Larochelle, “Optimization as a model for few-shot learning,” in ICLR, 2017.

[3] Jake Snell, Kevin Swersky, and Richard Zemel, “Prototypical networks for few-shot learning,” in NeurIPS, 2017, pp. 4077– 4087.

[4] Flood Sung, Yongxin Yang, Li Zhang, Tao Xiang, Philip HS Torr,andTimothyMHospedales, “Learningtocompare:Relationnet workforfew-shotlearning,” inCVPR,2018,pp.1199– 1208.

[5] Andrei A Rusu, Dushyant Rao, Jakub Sygnowski, Oriol Vinyals, Razvan Pascanu, Simon Osindero, and Raia Hadsell, “Meta- learning with latent embedding optimization,” arXiv preprint arXiv:1807.05960, 2018.

[6] Siyuan Qiao, Chenxi Liu, Wei Shen, and Alan L Yuille, “Fewshot image recognition by predicting parameters from activations,” in CVPR, 2018, pp. 7229–7238.

[7] XiangJiang,MohammadHavaei,FarshidVarno,GabrielChartra nd,NicolasChapados,andStanMatwin, “Learningtolearn with conditional class dependencies,” in ICLR, 2019.

[8] BorisOreshkin,PauRodr´ıguezL´opez,andAlexandreLacoste, “Tadam: Task dependent adaptive metric for improved fewshot learning,” in NeurIPS, 2018, pp. 719–729.

[9] Chelsea Finn, Pieter Abbeel, and Sergey Levine, “Modelagnostic meta-learning for fast adaptation of deep networks,” in ICML. JMLR, 2017, pp. 1126–1135.

[10] Gregory Koch, Richard Zemel, and Ruslan Salakhutdinov, “Siamese neural networks for one-shot image recognition,” in ICML Deep Learning Workshop, 2015, vol. 2.

[11] Philip Bachman, R Devon Hjelm, and William Buchwalter, “Learning representations by maximizing mutual information

across views,” arXiv preprint arXiv:1906.00910, 2019.

[12] Kaiming He, Xiangyu Zhang, Shaoqing Ren, and Jian Sun,“Deep residual learning for image recognition,” inCVPR. IEEE, 2016, pp. 770–778.

[13] Bernaschi M , Castiglione F , Ferranti A. ProtNet: A tool for stochastic simulations of protein interaction networks dynamics[J]. BMC Bioinformatics, 2007, 8 Suppl 1(Suppl 1):S4.

[14] CodingFish，自己搭建一个神经网络进行识别分类，https:// zhuanlan.zhihu.com/p/98981710,(2020.7.1).

[15] 杨卫红 . 数据库编程与图像处理 [J]. 电脑编程技巧与维护 , 2013(16):44-46.