更新2021/12/6
到目前为止看了不少主动学习的文献,简单做一下一些目前为止了解到的主动学习方法的整理吧。
起初是精读的文献中整理的,后来发现在精读文献的方法比较里也有一些比较经典的主动学习方法,附有略写名字和一些文献笔记,部分笔记可能因本人知识储备少存在错误。有一部分没有附上个人笔记就是还没有看,先存个档。
目录
1、A survey on active learning and human-in-the-loop deep learning for medical image analysis
1、core-set:ACTIVE LEARNING FOR CONVOLUTIONAL NEURAL NETWORKS: A CORE-SET APPROACH 核心集方法
2、MIAL:Multiple Instance Active Learning for Object Detection用于目标检测的多实例主动学习
3、CDAL:Contextual Diversity for Active Learning 主动学习的语境多样性
4、core-set GCN:Sequential Graph Convolutional Network for Active Learning 用于主动学习的序列图卷积神经网络
5、VAAL:Variational Adversarial Active Learning
6、Learning Loss:Learning loss for active learning
7、Feat Prop:Active Learning for Graph Neural Networksvia Node Feature Propagation
8、AQPL:Improving Model Robustness by Adaptively Correcting Perturbation Levels with Active Querie
9、MC-Dropout:Deep Bayesian Active Learning with Image Data
10、GAAL:Generative Adversarial Active Learning
11、CEAL:Cost-Effective Active Learning for Deep Image Classification
主动学习综述
1、A survey on active learning and human-in-the-loop deep learning for medical image analysis
医学图像分析中的主动学习和人在回路深度学习综述 https://arxiv.org/abs/1910.02923 这篇文献是我的主动学习启蒙
主动学习方法
有一些非常非常常见的方法(如果不是实在不懂的方法,一般我都只看2019后的文献)例如random、uncertain这类我就不列入了。
1、core-set:ACTIVE LEARNING FOR CONVOLUTIONAL NEURAL NETWORKS: A CORE-SET APPROACH 核心集方法
常见的方法core-set ,在很多文献里都有提及到,简单做过阅读笔记 core-set
2、MIAL:Multiple Instance Active Learning for Object Detection用于目标检测的多实例主动学习
cvpr2021的方法,我觉得是提出了一个非常好的背景不确定性和实例不确定性的观点吧 MIAL
3、CDAL:Contextual Diversity for Active Learning 主动学习的语境多样性
简单了解了一下,2020年的文献 CDAL
4、core-set GCN:Sequential Graph Convolutional Network for Active Learning 用于主动学习的序列图卷积神经网络
cvpr2021的文献,这里提到了一个我没有接触过的GCN core-setGCN
5、VAAL:Variational Adversarial Active Learning
1904.00370.pdf (arxiv.org) iccv2019 Variational Adversarial Active Learning_小鼠太郎的博客-CSDN博客
6、Learning Loss:Learning loss for active learning
cvpr2019,比较经典的设计求排列损失 learning loss
7、Feat Prop:Active Learning for Graph Neural Networksvia Node Feature Propagation
2019 1910.07567.pdf (arxiv.org)
8、AQPL:Improving Model Robustness by Adaptively Correcting Perturbation Levels with Active Querie
2021,AAAI 2103.14824.pdf (arxiv.org) 文献笔记
9、MC-Dropout:Deep Bayesian Active Learning with Image Data
2017,dropout Deep Bayesian Active Learning with Image Data (arxiv.org)
10、GAAL:Generative Adversarial Active Learning
首创性将gan加入到主动学习中,文献笔记
11、CEAL:Cost-Effective Active Learning for Deep Image Classification
16年 AL+CNN+分类
Cost-EffectivActive Learning forDeep Image Classification阅读笔记_小鼠太郎的博客-CSDN博客
待看文献补充:
1、An Efficient High-Quality Medical Lesion Image
Data Labeling Method Based on Active Learning 高效的一种关于医学图像的label方法
2、Task-Aware Variational Adversarial Active Learning 任务感知主动学习
3、Handling of uncertainty in medical data using machine
learning and probability theory techniques: a review
of 30 years (1991–2020) review:以往量化医学图像不确定性方法
4、A review of uncertainty quantification in deep learning: Techniques,
applications and challenges 关于不确定性量化的review