元学习研究综述

doi:10.11959/j.issn.1000-0801.2021009

摘要/Abstract

摘要：

深度学习和强化学习严重受限于小样本数据集，容易发生过拟合，无法实现类似于人类强泛化性的学习能力。元学习为此应运而生，以累积经验的方式形成“价值观”，基于本身的认知和价值判断能力对模型进行调整或优化，让智能体在实际环境中能快速学会各项复杂新任务，实现真正意义上的人工智能。首先概述了元学习的基本原理，然后根据其所采用的不同元知识形式，深入分析各类方法的研究现状，再探讨了元学习在少镜头学习、机器人学习和无监督学习等领域上的应用潜能，最后对其未来的发展趋势做出展望。

关键词: 小样本数据集, 强泛化性, 元学习, 人工智能

Abstract:

Deep learning and reinforcement learning are limited by small sample data set, which is impossible to realize the strong generalization learning ability.Meta-learning can make up for their shortcomings effectively.The values formed by accumulated experience feedback the corresponding signals to promote the model to adjust itself.It allows the artificial intelligence to learn to complete complex tasks quickly, which implements true artificial intelligence.Firstly, the basic principles of meta-learning were outlined.Secondly, according to the different forms of meta-knowledge, the research status of various methods was analyzed in depth.Finally, the application potential and the future development trends of meta-learning was discussed .

Key words: small sample data set, strong generalization, meta-learning, artificial intelligence

中图分类号:

TP391
TP18

朱应钊, 李嫚. 元学习研究综述[J]. 电信科学, 2021, 37(1): 22-31.

Yingzhao ZHU, Man LI. Review on meta-learning[J]. Telecommunications Science, 2021, 37(1): 22-31.

图/表 4

参考文献 63

[1]	KRIZHEVSKY A , SUTSKEVER I , HINTON G E . Imagenet classification with deep convolutional neural networks[C]// Proceedings of 26th Annual Conference on Neural Information Processing Systems. Cambridge:MIT Press, 2012: 1097-1105.
[2]	RUMELHART D E , HINTON G E , WILLIAMS R J . Learning representations by back propagating errors[J]. Nature, 1986,323(6088): 533-536.
[3]	GRAVES A , MOHAMED A , HINTON G . Speech recognition with deep recurrent neural networks[C]// Proceedings of the 38th IEEE International Conference on Acoustics,Speech,and Signal Processing. Piscataway:IEEE Press, 2013: 6645-6649.
[4]	BOTTOU L . From machine learning to machine reasoning[J]. Machine Learning, 2014,94(2): 133-149.
[5]	HINTON G E , OSINDERO S , TEH Y W . A fast learning algorithm for deep belief nets[J]. Neural Computation, 2006,18(7): 1527-1554.
[6]	龙慧, 朱定局, 田娟 . 深度学习在智能机器人中的应用研究综述[J]. 计算机科学, 2018,45(2): 43-47,52.
	LONG H , ZHU D J , TIAN J . Research on deep learning used in intelligent robots[J]. Computer Science, 2018,45(2): 43-47,52.
[7]	万里鹏, 兰旭光, 张翰博 ,等. 深度强化学习理论及其应用综述[J]. 模式识别与人工智能, 2019,32(1): 67-81.
	WANG L P , LAN X G , ZHANG H B ,et al. A review of deep reinforcement learning theory and application[J]. Pattern Recognition and Artificial Intelligence, 2019,32(1): 67-81.
[8]	李晨溪, 曹雷, 张永亮 ,等. 基于知识的深度强化学习研究综述[J]. 系统工程与电子技术, 2017,39(11): 2603-2613.
	LI C X , CAO X , ZHANG Y L ,et al. Knowledge-based deep reinforcement learning:a review[J]. Systems Engineering and Electronics, 2017,39(11): 2603-2613.
[9]	MATTEO H , JOSEPH M , HADO V H . Rainbow:combining improvements in deep reinforcement learning[C]// Proceedings of the 32nd AAAI Conference on Artificial Intelligence. Palo Alto:AAAI Press, 2018: 3215-3222.
[10]	ESHRATIFAR A , ABRISHAMI M , EIGEN D ,et al. A meta-learning approach for custom model training[C]// Proceedings of AAAI Conference on Artificial Intelligence. Palo Alto:AAAI Press, 2019: 9937-9938.
[11]	SIMONYAN K , ZISSERMAN A . Very deep convolutional networks for large-scale image recognition[C]// Proceedings of International Conference on Learning Representations, 2015.
[12]	HE K , ZHANG X , REN S ,et al. Deep residual learning for image recognition[C]// Proceedings of IEEE Conference on Computer Vision and Pattern Recognition. Piscataway:IEEE Press, 2016: 770-778.
[13]	SZEGEDY C , LIU W , JIA Y ,et al. Going deeper with convolutions[C]// Proceedings of IEEE Conference on Computer Vision and Pattern Recognition. Piscataway:IEEE Press, 2015: 1-9.
[14]	HINTON G , PLAUT D . Using fast weights to deblur old memories[C]// Proceedings of the 9th Annual Conference of the Cognitive Science Society. Hove:Psychology Press, 1987: 177-186.
[15]	FINN C , ABBEEL P , LEVINE S . Model-agnostic meta-learning for fast adaptation of deep networks[C]// Proceedings of International Conference on Machine Learning. New York:ACM Press, 2017: 1126-1135.
[16]	FINN C , RAJESWARAN A , KAKADE S ,et al. Online meta-learning[C]// Proceedings of International Conference on Machine Learning. New York:ACM Press, 2019: 1920-1930.
[17]	KINGMA D P , BA J . Adam:a method for stochastic optimization[C]// Proceedings of International Conference on Learning Representations. Palo Alto:AAAI Press, 2015: 1-15.
[18]	ANDRYCHOWICZ M , DENIL M , COLMENAREJO S G ,et al. Learning to learn by gradient descent by gradient descent[C]// Proceedings of Annual Conference on Neural Information Processing Systems. Cambridge:MIT Press, 2016: 3988-3996.
[19]	PARK E , REPIN D , SHEN L ,et al. Meta-curvature[C]// Proceedings of annual conference on neural information processing systems. Cambridge:MIT Press, 2019: 3309-3319.
[20]	HOUTHOOFT R , CHEN R Y , ISOLA P ,et al. Evolved policy gradients[C]// Proceedings of Annual Conference on Neural Information Processing Systems. Cambridge:MIT Press, 2018: 5405-5414.
[21]	LI Y Y , YANG Y X , ZHOU W ,et al. Feature-critic networks for heterogeneous domain generalization[C]// Proceedings of International Conference on Machine Learning. New York:ACM Press, 2019: 3915-3924.
[22]	SUNG F , YANG Y X , ZHANG L ,et al. Learning to compare:relation network for few-shot learning[C]// Proceedings of IEEE Conference on Computer Vision and Pattern Recognition. Piscataway:IEEE Press, 2018: 1199-1208.
[23]	KOSH G , ZEMEL R , SALAKHUTDINOV R . Siamese neural networks for one-shot image recognition[C]// Proceedings of International Conference on Machine Learning. New York:ACM Press, 2015: 6-36.
[24]	VINYALS O , BLUNDELL , LILLICRAP T ,et al. Matching networks for one shot learning[C]// Proceedings of Annual Conference on Neural Information Processing Systems. Cambridge:MIT Press, 2016: 3630-3638.
[25]	SHELL J , SWERSKY K , ZEMEL R S . Prototypical networks for few shot learning[C]// Proceedings of Annual Conference on Neural Information Processing Systems. Cambridge:MIT Press, 2017: 4077-4087.
[26]	GARCIA V , BRUNA J . Few-shot learning with graph neural networks[C]// Proceedings of International Conference on Learning Representations. Palo Alto:AAAI Press, 2018: 1-12.
[27]	REN M , LIAO R , FETAYA E ,et al. Incremental few-shot learning with attention attractor networks[C]// Proceedings of Annual Conference on Neural Information Processing Systems. Cambridge:MIT Press, 2019: 5276-5286.
[28]	HOU R , CHANG H , MA B P ,et al. Cross attention network for few-shot classification[C]// Proceedings of Annual Conference on Neural Information Processing Systems. Cambridge:MIT Press, 2019: 4005-4016.
[29]	FRANCESCHI L , FRASCONI P , SALZO S ,et al. Bilevel programming for hyperparameter optimization and meta-learning[C]// Proceedings of International Conference on Machine Learning. New York:ACM Press, 2018: 1563-1572.
[30]	LECUN Y , BOTTOU L , BENGIO ,et al. Gradient-based learning applied to document recognition[J]. Proceedings of the IEEE, 1998,86(11): 2278-2324.
[31]	ZOPH B , LE Q V . Neural architecture search with reinforcement learning[C]// Proceedings of International Conference on Machine Learning. New York:ACM Press, 2017: 459-468.
[32]	REAL E AGGARWAL , HUANG Y P ,et al. Regularized evolution for image classifier architecture search[C]// Proceedings of AAAI Conference on Artificial Intelligence. Palo Alto:AAAI Press, 2019: 4780-4789.
[33]	HOCHREITER S , YOUNGER A S , CONWELL P R ,et al. Learning to learn using gradient descent[C]// Proceedings of International Conference on Artificial Neural Networks. Berlin:Springer Press, 2001: 87-94.
[34]	SANTOTO A , BARTUNOV S , BOTVINICK M ,et al. Meta learning with memory-augmented neural networks[C]// Proceedings of International Conference on Machine Learning. New York:ACM Press, 2016: 1842-1850.
[35]	RAKELLY K , ZHOU A QUILLEN , FINN C ,et al. Efficient off-policy meta-reinforcement learning via probabilistic context variables[C]// Proceedings of International Conference on Machine Learning. New York:ACM Press, 2019: 5331-5340.
[36]	GIDARIS S , KOMODAKIS N . Dynamic few-shot visual learning without forgetting[C]// Proceedings of IEEE Conference on Computer Vision and Pattern Recognition. Piscataway:IEEE Press, 2018: 4367-4375.
[37]	REN M , LIAO R , FETAYA E ,et al. Incremental few-shot learning with attention attractor networks[C]// Proceedings of Annual Conference on Neural Information Processing Systems. Cambridge:MIT Press, 2019: 5276-5286.
[38]	KANG B , LIU Z , WANG X ,et al. Few-shot object detection via feature reweighting[C]// Proceedings of IEEE International Conference on Computer Vision. Piscataway:IEEE Press, 2019: 8419-8428.
[39]	MANUEL J , RUA P , ZHU X ,et al. Incremental few-shot object detection[C]// Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 2020.
[40]	GUO J Z , ZHU X Y , ZHAO C X ,et al. Learning meta face recognition in unseen domains[C]// Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 2020.
[41]	NGUYEN B D , DO T T , NGUYEN B X ,et al. Overcoming data limitation in medical visual question answering[C]// Proceedings of Medical Image Computing and Computer Assisted Intervention Society. Berlin:Springer Press, 2019: 522-530.
[42]	WANG T , LIU M , TAO A ,et al. Few-shot video-to-video synthesis[C]// Proceedings of Annual Conference on Neural Information Processing Systems. Cambridge:MIT Press, 2019: 5014-5025.
[43]	刘乃军, 鲁涛, 蔡莹皓 ,等. 机器人操作技能学习方法综述[J]. 自动化学报, 2019,45(3): 458-470.
	LIU N J , LU T , CAI Y H ,et al. A review of robot manipulation skills learning methods[J]. Acta Automatica Sinica, 2019,45(3): 458-470.
[44]	李帅龙, 张会文, 周维佳 . 模仿学习方法综述及其在机器人领域的应用[J]. 计算机工程与应用, 2019,55(4): 17-30.
	LI S L , ZHANG H W , ZHOU W J . Review of imitation learning methods and its application in robotics[J]. Computer Engineering and Applications, 2019,55(4): 17-30.
[45]	王薇, 吴锋, 周风余 . 机器人操作技能自主认知与学习的研究现状与发展趋势[J]. 山东大学学报, 2019,49(6): 11-24.
	WANG W , WU F , ZHOU F Y . Research status and development trend of autonomous cognition and learning of robot manipulation skills[J]. Journal of Shandong University ( Engineering Science) , 2019,49(6): 11-24.
[46]	DUAN Y , ANDRYCHOWICZ M , STADIE B C ,et al. One-shot imitation learning[C]// Proceedings of Annual Conference on Neural Information Processing Systems. Cambridge:MIT Press, 2017: 1087-1098.
[47]	FINN C , YU T H , ZHANG T H ,et al. One-shot visual imitation learning via meta-learning[C]// Proceedings of Annual Conference on Robot Learning. Cambridge:MIT Press, 2017: 357-368.
[48]	YU T H FINN C , XIE A ,et al. One-shot imitation from observing humans via domain-adaptive meta-learning[C]// Proceedings of International Conference on Learning Representations, 2018.
[49]	NAGABANDI A , CLAVERA I , LIU S ,et al. Learning to adapt in dynamic,real-world environments through meta-reinforcement learning[C]// Proceedings of International Conference on Learning Representations, 2019.
[50]	GARG V , KALAI A . Supervising unsupervised learning[C]// Proceedings of Annual Conference on Neural Information Processing Systems. Cambridge:MIT Press, 2018: 4996-5006.
[51]	ANTONIOU A , STORKEY A . Learning to learn by self- critique[C]// Proceedings of Annual Conference on Neural Information Processing Systems. Cambridge:MIT Press, 2019: 9936-9946.
[52]	KLEJCH O , FAINBERG J , BELL P . Learning to adapt:a meta-learning approach for speaker adaptation[C]// Proceedings of the 19th Annual Conference of the International Speech Communication Association, 2018.
[53]	HSU J Y , CHEN Y J , LEE H Y . Meta learning for end-to-end low-resource speech recognition[C]// Proceedings of International Conference on Acoustics,Speech and Signal Processing, 2019.
[54]	LI J , WONG Y , ZHAO Q ,et al. Learning to learn from noisy labeled data[C]// Proceedings of IEEE Conference on Computer Vision and Pattern Recognition. Piscataway:IEEE Press, 2019: 5051-5059.
[55]	SHU J , XIE Q , YI L ,et al. Meta-weight-net:learning an explicit mapping for sample weighting[C]// Proceedings of Annual Conference on Neural Information Processing Systems. Cambridge:MIT Press, 2019: 1917-1928.
[56]	CHEN S Y , WANG W Y , PAN S J . MetaQuant:learning to quantize by learning to penetrate non-differentiable quantization[C]// Proceedings of Annual Conference on Neural Information Processing Systems. Cambridge:MIT Press, 2019: 3918-3928.
[57]	BALAJI Y , SANKARANARAYANAN S , CHELLAPPA R . MetaReg:towards domain generalization using meta- regularization[C]// Proceedings of Annual Conference on Neural Information Processing Systems. Cambridge:MIT Press, 2018: 1006-1016.
[58]	LI D , YANG Y , SONG Y ,et al. Learning to generalize:meta-learning for domain generalization[C]// Proceedings of AAAI Conference on Artificial Intelligence. Palo Alto:AAAI Press, 2018: 3490-3497.
[59]	BROCK A , LIM T , RITCHIE J M ,et al. SMASH:one-shot model architecture search through hyper networks[C]// Proceedings of International Conference on Learning Representations.[S.l.:s.n.], 2018.
[60]	LEE K , MAJI S , RAVICHANDRAN A ,et al. Meta-learning with differentiable convex optimization[C]// Proceedings of IEEE Conference on Computer Vision and Pattern Recognition. Piscataway:IEEE Press, 2019: 10657-10665.
[61]	李茂莹, 杨柳, 胡清华 . 同构迁移学习理论和算法研究进展[J]. 南京信息工程大学学报(自然科学版), 2019,11(3): 269-277.
	LI M Y , YANG L , HU Q H . A survey on theories and algorithms about homogeneous transfer learning[J]. Journal of Nanjing University of Information Science and Technology (Natural Science Edition) , 2019,11(3): 269-277.
[62]	赵鹏, 高浩渊, 姚晟 ,等. 面向弱匹配的跨媒异构迁移学习[J]. 计算机辅助设计与图形学学报, 2019,31(11): 1963-1972.
	ZHAO P , GAO H Y , YAO S ,et al. Cross-media heterogeneous transfer learning oriented to semi-paired problem[J]. Journal of Computer-Aided Design ＆ Computer Graphics, 2019,31(11): 1963-1972.
[63]	朱应钊 . 异构迁移学习研究综述[J]. 电信科学, 2020,36(3): 100-110.
	ZHU Y Z . Review on heterogeneous transfer learning[J]. Telecommunications Science, 2020,36(3): 100-110.