面向小样本数据的机器学习方法研究综述

doi:10.19678/j.issn.1000-3428.0065347

摘要/Abstract

摘要： 小样本学习是面向小样本数据的机器学习，旨在利用较少的有监督样本数据去构建能够解决实际问题的机器学习模型。小样本学习能够解决传统机器学习方法在样本数据不充分时性能严重下降的问题，可以为新型小样本任务实现低成本和快速的模型部署，缩小人类智能与人工智能之间的距离，对推动发展通用型人工智能具有重要意义。从小样本学习的概念、基础模型和实际应用入手，系统梳理当前小样本学习的相关工作，将小样本学习方法分类为基于模型微调、基于数据增强、基于度量学习和基于元学习，并具体阐述这4大类方法的核心思想、基本模型、细分领域和最新研究进展，以及每一类方法在科学研究或实际应用中存在的问题，总结目前小样本学习研究的常用数据集和评价指标，整理基于部分典型小样本学习方法在Omniglot和Mini-ImageNet数据集上的实验结果。最后对各种小样本学习方法及其优缺点进行总结，分别从数据层面、理论研究和应用研究3个方面对小样本学习的未来研究方向进行展望。

关键词: 小样本学习, 小样本数据, 机器学习, 深度学习, 数据增强

Abstract: Few-shot learning is a type of machine learning method for small sample data that operates by using less supervised sample data to build machine learning models that can solve practical problems.Therefore, few-shot learning can be used to solve the serious performance degradation problem in traditional machine learning methods when a small sample data is used, and can achieve low-cost and rapid model deployment for new few-sample tasks, which has the potential of narrowing the distance between human intelligence and artificial intelligence and promote the general importance of artificial intelligence development.This paper systematically sorts out the existing related studies on few-shot learning and classifies the methods on few-shot learning into model-based fine-tuning, data augmentation, metric-based learning, and meta-learning based on the concept, basic model, and practical application of few-shot learning. Moreover, the core ideas, basic models, subdivision fields, and latest research progress in these four method categories are specifically expounded, and the problems existing in the scientific research and practical application of each method category are outlined.Data sets and evaluation indicators are also obtained, and the experimental results are organized based on typical few-shot learning methods with Omniglot and Mini-ImageNet datasets.Additionally, the advantages and disadvantages of various few-shot learning methods are summarized.Finally, data-level theoretical and applied research approaches, and potential future research directions of few-shot learning, are determined.

Key words: few-shot learning, small sample data, machine learning, deep learning, data augmentation

中图分类号:

TP391

陈良臣, 傅德印. 面向小样本数据的机器学习方法研究综述[J]. 计算机工程, 2022, 48(11): 1-13.

CHEN Liangchen, FU Deyin. Survey on Machine Learning Methods for Small Sample Data[J]. Computer Engineering, 2022, 48(11): 1-13.

https://www.ecice06.com/CN/Y2022/V48/I11/1

图/表 12

20230203202008

20230203202012

20230203202016

20230203202019

20230203202023

20230203202026

20230203202029

20230203202033

20230203202037

20230203202041

20230203202044

20230203202048

参考文献

[1] 赵凯琳, 靳小龙, 王元卓.小样本学习研究综述[J].软件学报, 2021, 32(2):349-369. ZHAO K L, JIN X L, WANG Y Z.Survey on few-shot learning[J].Journal of Software, 2021, 32(2):349-369.(in Chinese)
[2] SHU J, XU Z B, MENG D Y.Small sample learning in big data era[EB/OL].[2022-06-10].https://arxiv.org/abs/1808.04572.
[3] LI F F, FERGUS R, PERONA P.One-shot learning of object categories[J].IEEE Transactions on Pattern Analysis and Machine Intelligence, 2006, 28(4):594-611.
[4] 宋闯, 赵佳佳, 王康, 等.面向智能感知的小样本学习研究综述[J].航空学报, 2020, 41(S1):723-756. SONG C, ZHAO J J, WANG K, et al.A survey of few shot learning based on intelligent perception[J].Acta Aeronautica et Astronautica Sinica, 2020, 41(S1):723-756.(in Chinese)
[5] 葛轶洲, 刘恒, 王言, 等.小样本困境下的深度学习图像识别综述[J].软件学报, 2022, 33(1):193-210. GE Y Z, LIU H, WANG Y, et al.Survey on deep learning image recognition in dilemma of small samples[J].Journal of Software, 2022, 33(1):193-210.(in Chinese)
[6] KAISER L, NACHUM O, ROY A, et al.Learning to remember rare events[C]//Proceedings of IEEE ICLRʼ17.Washington D.C., USA:IEEE Press, 2017:1568-1577.
[7] DUAN Y, ANDRYCHOWICZ M, STADIE B C, et al.One-shot imitation learning[EB/OL].[2022-06-10].https://arxiv.org/abs/1703.07326.
[8] 刘嘉政.基于卷积神经网络的小样本树皮图像识别方法[J].西北林学院学报, 2019, 34(4):230-235. LIU J Z.Small sample bark image recognition method based on convolutional neural network[J].Journal of Northwest Forestry University, 2019, 34(4):230-235.(in Chinese)
[9] KANG B Y, LIU Z, WANG X, et al.Few-shot object detection via feature reweighting[C]//Proceedings of 2019 IEEE/CVF International Conference on Computer Vision.Washington D.C., USA:IEEE Press, 2019:8419-8428.
[10] PFISTER T, CHARLES J, ZISSERMAN A.Domain-adaptive discriminative one-shot learning of gestures[C]//Proceedings of IEEE ECCVʼ14.Washington D.C., USA:IEEE Press, 2014:814-829.
[11] 孙存威, 文畅, 谢凯, 等.深度迁移模型下的小样本声纹识别方法[J].计算机工程与设计, 2018, 43(12):3816-3822. SUN C W, WEN C, XIE K, et al.Voiceprint recognition method of small sample based on deep migration model[J].Computer Engineering and Design, 2018, 43(12):3816-3822.(in Chinese)
[12] 程林, 袁慊, 王瑜, 等.自体支气管基底层细胞治疗慢性阻塞性肺疾病的小样本探索性研究[J].重庆医学, 2019, 48(23):4012-4016. CHENG L, YUAN Q, WANG Y, et al.A small sample exploratory study of autogenous bronchial basal cells for the treatment of chronic obstructive pulmonary disease[J].Chongqing Medical, 2019, 48(23):4012-4016.(in Chinese)
[13] 何喜军, 马珊, 武玉英, 等.小样本下多维指标融合的电商产品销量预测[J].计算机工程与应用, 2019, 55(15):177-184. HE X J, MA S, WU Y Y, et al.E-commerce product sales forecast with multi-dimensional index integration under small sample[J].Computer Engineering and Applications, 2019, 55(15):177-184.(in Chinese)
[14] 陈龙, 张峰, 蒋升.小样本条件下基于深度森林学习模型的典型军事目标识别方法[J].中国电子科学研究院学报, 2019, 14(3):232-237. CHEN L, ZHANG F, JIANG S.Deep forest learning for military object recognition under small training set condition[J].Journal of Chinese Academy of Electronics, 2019, 14(3):232-237.(in Chinese)
[15] 王佳林.基于深度学习的小样本入侵检测研究[D].北京:北京交通大学, 2020. WANG J L.Research on intrusion detection of small samples based on deep learning[D].Beijing:Beijing Jiaotong University, 2020.(in Chinese)
[16] LI F F, PERONA F R.A Bayesian approach to unsupervised one-shot learning of object categories[C]//Proceedings of the 9th IEEE International Conference on Computer Vision.Washington D.C., USA:IEEE Press, 2019:1134-1141.
[17] MITCHELL T M.Machine learning[M].[S.1.]:McGraw-Hill, Inc., 1997.
[18] 吴章凯.基于特征增强元学习优化的小样本学习的研究[D].南京:南京大学, 2020. WU Z K.A research on few-shot learning based on feature enhanced meta-learning optimization[D].Nanjing:Nanjing University, 2020.(in Chinese)
[19] DHILLON G S, CHAUDHARI P, RAVICHANDRAN A, et al.A baseline for few-shot image classification[C]//Proceedings of IEEE ICLRʼ20.Washington D.C., USA:IEEE Press, 2020:2542-1556.
[20] WANG X, HUANG T E, DARRELL T, et al.Frustratingly simple few-shot object detection[EB/OL].[2022-06-10].https://arxiv.org/abs/2003.06957.
[21] HOWARD J, RUDER S.Universal language model fine-tuning for text classification[C]//Proceedings of the 56th Annual Meeting of the Association for Computational Linguistics.Stroudsburg, USA:Association for Computational Linguistics, 2018:3517-3529.
[22] NAKAMURA A, HARADA T.Revisiting fine-tuning for few-shot learning[EB/OL].[2022-06-10].https://arxiv.org/abs/1910.00216.
[23] LIU H K, TAM D, MUQEETH M, et al.Few-shot parameter-efficient fine-tuning is better and cheaper than in-context learning[EB/OL].[2022-06-10].https://arxiv.org/abs/2205.05638.
[24] HSU K, LEVINE S, FINN C.Unsupervised learning via meta-learning[C]//Proceedings of IEEE International Conference on Learning Representations.Washington D.C., USA:IEEE Press, 2019:235-246.
[25] JI Z L, ZOU X L, HUANG T J, et al.Unsupervised few-shot learning via self-supervised training[EB/OL].[2022-06-10].https://arxiv.org/abs/1912.12178.
[26] KHODADADEH S, BLNI L, SHAH M.Unsupervised meta-learning for few-shot image classification[C]//Proceedings of NIPSʼ19.Cambridge, USA:MIT Press, 2019:233-246.
[27] ANTONIOU A, STORKEY A.Assume, augment and learn:unsupervised few-shot meta-learning via random labels and data augmentation[EB/OL].[2022-06-10].https://arxiv.org/abs/1902.09884.
[28] WANG Y K, XU C M, LIU C, et al.Instance credibility inference for few-shot learning[C]//Proceedings of 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition.Washington D.C., USA:IEEE Press, 2020:12833-12842.
[29] 秦铁鑫.基于数据扩充的小样本学习算法研究[D].南京:南京大学, 2021. QIN T X.The study of data augmentation based few-shot learning algorithms[D].Nanjing:Nanjing University, 2021.(in Chinese)
[30] QIN T, LI W, SHI Y, et al.Diversity helps:unsupervised few-shot learning via distribution shift-based data augmentation[EB/OL].[2022-06-10].https://arxiv.org/abs/2004.05805.
[31] LI P, ZHAO G P, XU X H.Coarse-to-fine few-shot classification with deep metric learning[J].Information Sciences, 2022, 610:592-604.
[32] REN M Y, TRIANTAFILLOU E, RAVI S, et al.Meta-learning for semi-supervised few-shot classification[EB/OL].[2022-06-10].https://arxiv.org/abs/1803.00676.
[33] LIU Y, LEE J, PARK M, et al.Learning to propagate labels:transductive propagation network for few-shot learning[EB/OL].[2022-06-10].https://arxiv.org/abs/1805.10002.
[34] WANG Y X, HEBERT M.Learning from small sample sets by combining unsupervised meta-training with CNNs[C]//Proceedings of NIPSʼ19.Cambridge, USA:MIT Press, 2016:244-252.
[35] BONEY R, ILIN A.Semi-supervised few-shot learning with MAML[C]//Proceedings of ICLRʼ18.Washington D.C., USA:IEEE Press, 2018:568-579.
[36] HOU R B, CHANG H, MA B P, et al.Cross attention network for few-shot classification[EB/OL].[2022-06-10].https://arxiv.org/abs/1910.07677.
[37] GOODFELLOW I, POUGET-ABADIE J, MIRZA M, et al.Generative adversarial nets[C]//Proceedings of NIPSʼ14.Cambridge, USA:MIT Press, 2014:2672-2680.
[38] ANTONIOU A, STORKEY A, EDWARDS H.Data augmentation generative adversarial networks[EB/OL].[2022-06-10].https://arxiv.org/abs/1711.04340.
[39] MEHROTRA A, DUKKIPATI A.Generative adversarial residual pairwise networks for one shot learning[EB/OL].[2022-06-10].https://arxiv.org/abs/1703.08033.
[40] CHEN Z T, FU Y W, WANG Y X, et al.Image deformation meta-networks for one-shot learning[C]//Proceedings of 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition.Washington D.C., USA:IEEE Press, 2019:8672-8681.
[41] WANG Y X, GIRSHICK R, HEBERT M, et al.Low-shot learning from imaginary data[C]//Proceedings of 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition.Washington D.C., USA:IEEE Press, 2018:7278-7286.
[42] ZHANG R, CHE T, GHAHRAMANI Z, et al.Metagan:an adversarial approach to few-shot learning[C]//Proceedings of NIPSʼ19.Cambridge, USA:MIT Press 2018:2365-2374.
[43] SCHWARTZ E, KARLINSKY L, SHTOK J, et al.Delta-encoder:an effective sample synthesis method for few-shot object recognition[C]//Proceedings of NIPSʼ18.Cambridge, USA:MIT Press, 2018:2845-2855P.
[44] XIAN Y Q, SHARMA S, SCHIELE B, et al.F-VAEGAN-D2:a feature generating framework for any-shot learning[C]//Proceedings of 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition.Washington D.C., USA:IEEE Press, 2019:10267-10276.
[45] CHEN W J.Semi-supervised learning study summary[J].Computer Knowledge and Technology, 2011, 7(16):3887-3889.
[46] SUBEDI B, SATHISHKUMAR V E, MAHESHWARI V, et al.Feature learning-based generative adversarial network data augmentation for class-based few-shot learning[J].Mathematical Problems in Engineering, 2022, 22:1-20.
[47] DIXIT M, KWITT R, NIETHAMMER M, et al.AGA:attribute-guided augmentation[C]//Proceedings of 2017 IEEE Conference on Computer Vision and Pattern Recognition.Washington D.C., USA:IEEE Press, 2017:3328-3336.
[48] HARIHARAN B, GIRSHICK R.Low-shot visual recognition by shrinking and hallucinating features[C]//Proceedings of 2017 IEEE International Conference on Computer Vision.Washington D.C., USA:IEEE Press, 2017:3037-3046.
[49] LIU B, WANG X D, DIXIT M, et al.Feature space transfer for data augmentation[C]//Proceedings of 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition.Washington D.C., USA:IEEE Press, 2018:9090-9098.
[50] CHEN Z T, FU Y W, ZHANG Y D, et al.Multi-level semantic feature augmentation for one-shot learning[EB/OL].[2022-06-10].https://arxiv.org/abs/1804.05298.
[51] ZHANG H G, ZHANG J, KONIUSZ P.Few-shot learning via saliency-guided hallucination of samples[C]//Proceedings of 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition.Washington D.C., USA:IEEE Press, 2019:2765-2774.
[52] 周琳钧.小样本学习理论方法研究[D].北京:清华大学, 2021. ZHOU L J.Few-shot learning:theories and methods[D].Beijing:Tsinghua University, 2021.(in Chinese)
[53] TSENG H Y, LEE H Y, HUANG J B, et al.Cross-domain few-shot classification via learned feature-wise transformation[EB/OL].[2022-06-10].https://arxiv.org/abs/2001.08735.
[54] JING K L, ZHANG X M, YANG Z Y, et al.Feature augmentation learning for few-shot palmprint image recognition with unconstrained acquisition[C]//Proceedings of IEEE International Conference on Acoustics, Speech and Signal Processing.Washington D.C., USA:IEEE Press, 2022:3323-3327.
[55] AURELIEN B, AMAURY H, MARC S.A survey on metric learning for feature vectors andstructured data[EB/OL].[2022-06-10].https://arxiv.org/abs/1306.6709.
[56] SNELL J, SWERSKY K, ZEMEL R S.Prototypical networks for few-shot learning[C]//Proceedings of NIPSʼ17.Cambridge, USA:MIT Press, 2017:4077-4087.
[57] SUNG F, YANG Y X, ZHANG L, et al.Learning to compare:relation network for few-shot learning[C]//Proceedings of 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition.Washington D.C., USA:IEEE Press, 2018:1199-1208.
[58] LI W B, WANG L, XU J L, et al.Revisiting local descriptor based image-to-class measure for few-shot learning[C]//Proceedings of 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition.Washington D.C., USA:IEEE Press, 2019:7253-7260.
[59] LI W B, XU J L, HUO J, et al.Distribution consistency based covariance metric networks for few-shot learning[C]//Proceedings of AAAI Conference on Artificial Intelligence.[S.1.]:AAAI Press, 2019:8642-8649.
[60] VINYALS O, BLUNDELL C, LILLICRAP T, et al.Matching networks for one shot learning[EB/OL].[2022-06-10].https://arxiv.org/abs/1606.04080.
[61] HAN G X, HUANG S Y, MA J W, et al.Meta faster R-CNN:towards accurate few-shot object detection with attentive feature alignment[C]//Proceedings of AAAI Conference on Artificial Intelligence.[S.1.]:AAAI Press, 2022:780-789.
[62] ZHANG J H, MAIMAITI M, XING G, et al.MGIMN:multi-grained interactive matching network for few-shot text classification[EB/OL].[2022-06-10].https://arxiv.org/abs/2204.04952.
[63] LIU B, CAO Y, LIN Y T, et al.Negative margin matters:understanding margin in few-shot classification[EB/OL].[2022-06-10].https://arxiv.org/abs/2003.12060.
[64] 孙牧野.基于度量学习的小样本图像分类技术研究[D].西安:西安电子科技大学, 2020. SUN M Y.Research on few shot learning for image classification based on metric learning[D].Xi'an:Xidian University, 2020.(in Chinese)
[65] CHAI Y H, DU L, QIU J, et al.Dynamic prototype network based on sample adaptation for few-shot malware detection[J].IEEE Transactions on Knowledge and Data Engineering, 2022, 99(35):1-10.
[66] SUNG F, YANG Y X, ZHANG L, et al.Learning to compare:relation network for few-shot learning[C]//Proceedings of 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition.Washington D.C., USA:IEEE Press, 2018:1199-1208.
[67] 安晨, 汪成亮, 廖超, 等.基于注意力关系网络的无线胶囊内镜图像分类方法[J].计算机工程, 2021, 47(10):252-259, 268. AN C, WANG C L, LIAO C, et al.Wireless capsule endoscopy image classification method based on attention relational network[J].Computer Engineering, 2021, 47(10):252-259, 268.(in Chinese)
[68] ABDELAZIZ M, ZHANG Z P.Multi-scale kronecker-product relation networks for few-shot learning[J].Multimedia Tools and Applications, 2022, 81(5):6703-6722.
[69] GARCIA V, BRUNA J.Few-shot learning with graph neural networks[EB/OL].[2022-06-10].https://arxiv.org/abs/1711.04043.
[70] 袁自勇, 高曙, 曹姣, 等.基于异构图卷积网络的小样本短文本分类方法[J].计算机工程, 2021, 47(12):87-94. YUAN Z Y, GAO S, CAO J, et al.Method for few-shot short text classification based on heterogeneous graph convolutional network[J].Computer Engineering, 2021, 47(12):87-94.(in Chinese)
[71] TAO X Y, HONG X P, CHANG X Y, et al.Few-shot class-incremental learning[C]//Proceedings of 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition.Washington D.C., USA:IEEE Press, 2020:12180-12189.
[72] YU T Y, HE S, SONG Y Z, et al.Hybrid graph neural networks for few-shot learning[C]//Proceedings of AAAI Conference on Artificial Intelligence.[S.1.]:AAAI Press, 2022:3179-3187.
[73] ZHAO K K, ZHANG Z Y, JIANG B, et al.LGLNN:label guided graph learning-neural network for few-shot learning[J].Neural Networks, 2022, 155:50-57.
[74] 刘鑫, 周凯锐, 何玉琳, 等.基于度量的小样本分类方法研究综述[J].模式识别与人工智能, 2021, 34(10):909-923. LIU X, ZHOU K R, HE Y L, et al.Survey of metric-based few-shot classification[J].Pattern Recognition and Artificial Intelligence, 2021, 34(10):909-923.(in Chinese)
[75] BENGIO, YOSHUA.A meta-transfer objective for learning to disentangle causal mechanisms[C]//Proceedings of IEEE ICLRʼ20.Washington D.C., USA:IEEE Press, 2020:356-368.
[76] FINN C, ABBEEL P, LEVINE S.Model-agnostic meta-learning for fast adaptation of deep net-works[C]//Proceedings of IEEE ICML ʼ20.Washington D.C., USA:IEEE Press, 2017:1126-1135.
[77] JAMAL M A, QI G J.Task agnostic meta-learning for few-shot learning[C]//Proceedings of 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition.Washington D.C., USA:IEEE Press, 2019:11711-11719.
[78] JIANG X, HAVAEI M, CHARTRAND G, et al.On the importance of attention in meta-learning for few-shot text classification[EB/OL].[2022-06-10].https://arxiv.org/abs/1806.00852.
[79] LI Z G, ZHOU F W, CHEN F, et al.Meta-SGD:learning to learn quickly for few-shot learning[EB/OL].[2022-06-10].https://arxiv.org/abs/1707.09835.
[80] RUSU A A, RAO D, SYGNOWSKI J, et al.Meta-learning with latent embedding optimization[EB/OL].[2022-06-10].https://arxiv.org/abs/1807.05960.
[81] LIU G D, WANG T L, ZHANG S X.Generating pseudo-labels adaptively for few-shot model-agnostic meta-learning[EB/OL].[2022-06-10].https://arxiv.org/abs/2207.0421.
[82] LOPEZ-PAZ D, RANZATO M.Gradient episodic memory for continual learning[EB/OL].[2022-06-10].https://arxiv.org/abs/1706.08840.
[83] SOH J W, CHO S, CHO N I.Meta-transfer learning for zero-shot super-resolution[C]//Proceedings of 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition.Washington D.C., USA:IEEE Press, 2020:3513-3522.
[84] LEI S Z, DONG B H, SHAN A K, et al.Attention meta-transfer learning approach for few-shot iris recognition[J].Computers and Electrical Engineering, 2022, 99:107848.
[85] RAVI S, LAROCHELLE H.Optimization as a model for few-shot learning[C]//Proceedings of the 5th International Conference on Learning Representations.Washington D.C., USA:IEEE Press, 2016:458-469.
[86] SANTORO A, BARTUNOV S, BOTVINICK M, et al.One-shot learning with memory-augmented neural networks[EB/OL].[2022-06-10].https://arxiv.org/abs/1605.06065.
[87] 刘颖, 雷研博, 范九伦, 等.基于小样本学习的图像分类技术综述[J].自动化学报, 2021, 47(2):297-315. LIU Y, LEI Y B, FAN J L, et al.Survey on image classification technology based on small sample learning[J].Acta Automatica Sinica, 2021, 47(2):297-315.(in Chinese)
[88] 李凡长, 刘洋, 吴鹏翔, 等.元学习研究综述[J].计算机学报, 2021, 44(2):422-446. LI F Z, LIU Y, WU P X, et al.A survey on recent advances in meta-learning[J].Chinese Journal of Computers, 2021, 44(2):422-446.(in Chinese)

选择文件类型/文献管理软件名称

选择包含的内容