计算机工程与应用 ›› 2021, Vol. 57 ›› Issue (7): 22-33.DOI: 10.3778/j.issn.1002-8331.2012-0200

• 热点与综述 • 上一篇    下一篇

深度神经网络的小样本学习综述

祝钧桃,姚光乐,张葛祥,李军,杨强,王胜,叶绍泽   

  1. 1.成都理工大学 信息科学与技术学院,成都 610059
    2.成都理工大学 人工智能研究中心,成都 610059
    3.成都理工大学 环境与土木工程学院,成都 610059
    4.成都理工大学 地质灾害防治与地质环境保护国家重点实验室,成都 610059
    5.深圳市勘察研究院有限公司,广东 深圳 518026
  • 出版日期:2021-04-01 发布日期:2021-04-02

Survey of Few Shot Learning of Deep Neural Network

ZHU Juntao, YAO Guangle, ZHANG Gexiang, LI Jun, YANG Qiang, WANG Sheng, YE Shaoze   

  1. 1.School of Information Science and Technology, Chengdu University of Technology, Chengdu 610059, China
    2.Research Center for Artificial Intelligence, Chengdu University of Technology, Chengdu 610059, China
    3.School of Environment and Civil Engineering, Chengdu University of Technology, Chengdu 610059, China
    4.State Key Laboratory of Geohazard Prevention and Geoenvironment Protection, Chengdu University of Technology, Chengdu 610059, China
    5.Shenzhen Investigation and Research Institute Co. , Ltd., Shenzhen, Guangdong 518026, China
  • Online:2021-04-01 Published:2021-04-02

摘要:

随着最近深度学习技术的蓬勃发展,深度神经网络(DNN)在大规模的图像分类与识别任务中取得了突破性的进展,但其在解决小样本学习问题时仍面临巨大挑战。小样本学习(FSL)是指在少量有监督样本的情况下学习一个能解决实际问题的模型,在深度学习领域具有重要意义。这促使该系统梳理了已有的DNN下的小样本学习工作,根据它们在解决小样本学习问题时所采用的技术,将DNN下的小样本学习解决方案分为四种策略:数据增强、度量学习、外部记忆、参数优化。根据这些策略,对现有的DNN下的小样本学习方法进行了全面的综述,同时总结了每一种策略在相关基准上的表现。强调了现有技术存在的局限性并对其未来的发展方向进行了展望,为今后的研究工作提供参考。

关键词: 小样本学习, 度量学习, 数据增强, 元学习, 深度神经网络

Abstract:

With the recent vigorous development of deep learning, Deep Neural Networks(DNN) have made exciting breakthrough in large-scale image classification and recognition tasks, but they still face huge challenges in solving few shot learning problems. Few Shot Learning(FSL) is defined as learning a model that can solve practical problems with a small number of supervised samples, which is of great significance in the field of deep learning. This prompts people to systematically combs the recent work of few shot learning of DNN, and divide the solution into four strategies according to the technology they used to solve the small sample learning problem:data augmentation, metric learning, external memory, parameter optimization. According to these strategies, it comprehensively reviews the existing few shot learning methods of DNN, and summarizes the performance of each strategy on relevant benchmarks. Finally, the limitations of the existing technology are emphasized and its future development direction is prospected to provide reference for future research work.

Key words: few shot learning, metric learning, data augmentation, meta learning, deep neural network