8 minute read

识别从未见过的数据类别，使模型具有知识迁移的能力；

zero shot learning

1 综述

  1. Learning to Detect Unseen Object Class by Between-Class Attribute Transfer
    CVPR 2009 2009 paper
    最早提出零样本学习；
    $\bullet \bullet$ detect

  2. Zero Shot Learning with Semantic Output Codes
    NIPS 2009 2009 paper
    $\bullet \bullet$ Output

  3. Recent Advances in Zero-shot Recognition
    2017 paper
    $\bullet \bullet$ advance

  4. A Survey of Zero-Shot Learning: Settings, Methods, and Applications
    TIST 2019 2019 paper | blog
    $\bullet \bullet$ Survey

2 理论

  1. Learning from one example through shared densities on transforms
    2000 paper
    $\bullet \bullet$
    贝叶斯单次学习算法的方法

  2. A Neural-Symbolic Architecture for Inverse Graphics Improved by Lifelong Meta-Learning
    2019-05-22 paper

  3. Embedded Meta-Learning: Toward more flexible deep-learning models
    2019-05-23 paper

3 小样本学习

3.1 基于度量的

  1. Prototypical Networks for Few-shot Learning
    NIPS 2017 2017-03-15 paper | openreview

3.2 基于模型的

  1. Task-Agnostic Meta-Learning for Few-shotLearning
    2018-05-20 paper | reddit

3.3 基于优化的

  1. Optimization as a model for few-shot learning
    ICLR 2017 paper | openreview | iclr_notes

  2. Learning to Compare: Relation Network for Few-Shot Learning
    CVPR 2018 2017-11-16 paper | pytorch
    文章探讨了这个为什么会有论文中的 idea；提出了一种基于度量（metric-based）的方法，只不过不是人为度量，而是让网络去度量；

3.4 其他

  1. Attentive Task-Agnostic Meta-Learning for Few-Shot Text Classification
    paper | openreview

  2. Learning Classifier Synthesis for Generalized Few-Shot Learning
    2019-06-07 paper

  3. Meta-Learning with Domain Adaptation for Few-Shot Learning under Domain Shift
    ICLR 2019 paper | openreview

  4. Learning to Propagate Labels: Transductive Propagation Network for Few-shot Learning
    ICLR 2019 2018-05-25 paper | tensorflow-official | pytorch-official | openreview | blog

  5. Adaptive Posterior Learning: few-shot learning with a surprise-based memory module
    ICLR 2019 2019-02-07 paper | openreview pytorch

  6. A Transductive Multi-Head Model for Cross-Domain Few-Shot Learning
    2020-06-08 paper

4 单样本学习

4.1 基于度量的

  1. Matching Networks for One Shot Learning
    NIPS 2016 2016-06-13 paper

4.2 其他

  1. One-shot learning of object categories
    2006 paper

5 零样本学习

  1. Zero-data learning of new tasks
    AAAI 2008 2008 paper

  2. Zero-Shot Learning with Semantic Output Codes
    NIPS 2009 2009 paper

  3. DeViSE: A Deep Visual-Semantic Embedding Model
    NIPS 2013 2013 paper
    DeViSE

  4. Zero-shot learning through cross-modal transfer
    NIPS 2013 2013-01-16 paper | matlab
    CMT

  5. Attribute-Based Classification for Zero-Shot Visual Object Categorization
    TPAMI 2013 2013 paper
    DAP

  6. Label Embedding for Attribute-Based Classification
    CVPR 2013 2013 paper
    ALE

  7. Transductive Multi-view Embedding for Zero-Shot Recognition and Annotation
    ECCV 2014 2014 paper | matlab
    TMV-BLP

  8. Zero-Shot Learning via Visual Abstraction
    ECCV 2014 2014 paper | matlab | project

  9. Zero-Shot Recognition with Unreliable Attributes
    NIPS 2014 2014 paper

  10. Evaluation of Output Embeddings for Fine-Grained Image Classification
    CVPR 2015 2015 paper | code
    $\bullet \bullet$ - SJE-II

  11. Zero-Shot Object Recognition by Semantic Manifold Distance
    CVPR 2015 2015 paper

  12. Zero-Shot Learning via Semantic Similarity Embedding
    ICCV 2015 2015 paper | code
    $\bullet \bullet$ - SSE

  13. Semi-Supervised Zero-Shot Classification with Label Representation Learning
    ICCV 2015 2015 paper
    LRL

  14. Unsupervised Domain Adaptation for Zero-Shot Learning
    ICCV 2015 2015 paper
    UDA

  15. Predicting Deep Zero-Shot Convolutional Neural Networks using Textual Descriptions
    ICCV 2015 2015 paper

  16. Transductive Multi-view Zero-Shot Learning
    TPAMI 2015 2015 paper | matlab
    TMV

  17. An embarrassingly simple approach to zero-shot learning
    ICML 2015 2015 paper | python
    $\bullet \bullet \bullet$ - EsZSL
    入门的工程；

  18. Label-Embedding for Image Classification
    TPAMI 2016 2016 paper
    ALE

  19. Relational Knowledge Transfer for Zero-Shot Learning
    AAAI 2016 2016 paper
    RKT

  20. An Empirical Study and Analysis of Generalized Zero-Shot Learning for Object Recognition in the Wild
    ECCV 2016 2016 paper

  21. Multi-Task Zero-Shot Action Recognition with Prioritised Data Augmentation
    ECCV 2016 2016 paper MTE

  22. Zero-Shot Recognition via Structured Prediction
    ECCV 2016 2016 paper

  23. Improving Semantic Embedding Consistency by Metric Learning for Zero-Shot Classification
    ECCV 2016 2016 paper

  24. Multi-Cue Zero-Shot Learning With Strong Supervision
    CVPR 2016 2016 paper | project MC-ZSL

  25. Latent Embeddings for Zero-Shot Classification
    CVPR 2016 2016 paper | code
    $\bullet \bullet$ - LATEM

  26. Less Is More: Zero-Shot Learning From Online Textual Documents With Noise Suppression
    CVPR 2016 2016 paper
    LIM

  27. Synthesized Classifiers for Zero-Shot Learning
    CVPR 2016 2016 paper | code-official | matlab-official
    $\bullet \bullet$ - SYNC

  28. Recovering the Missing Link: Predicting Class-Attribute Associations for Unsupervised Zero-Shot Learning
    CVPR 2016 2016 paper
    RML

  29. Zero-Shot Learning via Joint Latent Similarity Embedding
    CVPR 2016 2016 paper | code
    $\bullet \bullet$ - SLE

  30. Semantically Consistent Regularization for Zero-Shot Recognition
    CVPR 2017 2017 paper | caffe
    $\bullet \bullet$ - Deep-SCoRe

  31. Learning a Deep Embedding Model for Zero-Shot Learning
    CVPR 2017 2017 paper | tensorflow
    $\bullet \bullet$ - DEM

  32. From Zero-Shot Learning to Conventional Supervised Classification: Unseen Visual Data Synthesis
    CVPR 2017 2017 paper
    VDS

  33. Low-Rank Embedded Ensemble Semantic Dictionary for Zero-Shot Learning
    CVPR 2017 2017 paper
    ESD

  34. Semantic Autoencoder for Zero-Shot Learning
    CVPR 2017 2017 paper | matlab-official | pytorch
    $\bullet \bullet$ - SAE

  35. Zero-Shot Recognition Using Dual Visual-Semantic Mapping Paths
    CVPR 2017 2017 paper
    DVSM

  36. Matrix Tri-Factorization With Manifold Regularizations for Zero-Shot Learning
    CVPR 2017 2017 paper
    MTF-MR

  37. Gaze Embeddings for Zero-Shot Image Classification
    CVPR 2017 2017 paper | python
    $\bullet \bullet$ - SJE

  38. Zero-Shot learning - The Good, the Bad and the Ugly
    CVPR 2017 2017 paper | caffe
    $\bullet \bullet$ - GUB

  39. Attributes2Classname: A Discriminative Model for Attribute-Based Unsupervised Zero-Shot Learning
    ICCV 2017 2017 paper | tensorflow
    $\bullet \bullet$ - A2C

  40. Predicting Visual Exemplars of Unseen Classes for Zero-Shot Learning
    ICCV 2017 2017 paper | matlab
    PVE

  41. Learning Discriminative Latent Attributes for Zero-Shot Classification
    ICCV 2017 2017 paper
    LDL

  42. Zero-Shot Recognition via Direct Classifier Learning with Transferred Samples and Pseudo Labels
    AAAI 2017 2017 paper
    DCL

  43. Zero-Shot Learning via Category-Specific Visual-Semantic Mapping and Label Refinement
    2018 paper
    AEZSL

  44. Zero-Shot Dialog Generation with Cross-Domain Latent Actions
    SIGDIAL 2018 2018 paper | pytorch
    $\bullet \bullet$ - ZSGD

  45. Adversarial Zero-shot Learning With Semantic Augmentation
    AAAI 2018 2018 paper
    GANZrl

  46. Joint Dictionaries for Zero-Shot Learning
    AAAI 2018 2018 paper
    JDZsL

  47. Zero-Shot Learning via Class-Conditioned Deep Generative Models
    AAAI 2018 2018 paper
    VZSL

  48. Zero-Shot Learning With Attribute Selection
    AAAI 2018 2018 paper
    AS

  49. Deep Semantic Structural Constraints for Zero-Shot Learning
    AAAI 2018 2018 paper
    DSSC

  50. Towards Affordable Semantic Searching: Zero-Shot Retrieval via Dominant Attributes
    AAAI 2018 2018 paper
    ZsRDA

  51. Zero-shot learning-A comprehensive evaluation of the good, the bad and the ugly
    TPAMI 2018 2017-07-03 paper | project
    C-GUB

  52. Transductive Unbiased Embedding for Zero-Shot Learning
    CVPR 2018 2018-03-30 阿里 paper | blog
    TUE

  53. Zero-shot Recognition via Semantic Embeddings and Knowledge Graphs
    CVPR 2018 2018 paper | tensorflow
    $\bullet \bullet$ - GCN

  54. Preserving Semantic Relations for Zero-Shot Learning
    CVPR 2018 2018 paper
    PSR

  55. A Generative Adversarial Approach for Zero-Shot Learning From Noisy Texts
    CVPR 2018 2018 paper
    GAN-NT

  56. Zero-Shot Visual Recognition Using Semantics-Preserving Adversarial Embedding Networks
    CVPR 2018 2018 paper | tensorflow
    $\bullet \bullet$ - SP-AEN

  57. Multi-Label Zero-Shot Learning With Structured Knowledge Graphs
    CVPR 2018 2018 paper | project
    ML-SKG

  58. Generalized Zero-Shot Learning via Synthesized Examples
    CVPR 2018 2018 paper
    GZSL-SE

  59. Feature Generating Networks for Zero-Shot Learning
    CVPR 2018 2018 paper | code | project
    $\bullet \bullet$ - FGN

  60. Discriminative Learning of Latent Features for Zero-Shot Recognition
    CVPR 2018 2018-03-18 paper | blog
    $\bullet \bullet$ latent feature
    LDF

  61. Webly Supervised Learning Meets Zero-shot Learning: A Hybrid Approach for Fine-grained Classification
    CVPR 2018 2018 paper WSL

  62. Selective Zero-Shot Classification with Augmented Attributes
    ECCV 2018 2018 paper
    SZSL

  63. Learning Class Prototypes via Structure Alignment for Zero-Shot Recognition
    ECCV 2018 2018 paper
    LCP-SA

  64. Multi-modal Cycle-consistent Generalized Zero-Shot Learning
    ECCV 2018 2018 paper | tensorflow
    $\bullet \bullet$ - MC-ZSL

  65. Generalized Zero-Shot Learning with Deep Calibration Network
    NIPS 2018 2018 paper
    DCN

  66. Stacked Semantics-Guided Attention Model for Fine-Grained Zero-Shot Learning
    NIPS 2018 2018 paper S2GA

  67. Domain-Invariant Projection Learning for Zero-Shot Recognition
    NIPS 2018 2018 paper DIPL

  68. Generalized Zero- and Few-Shot Learning via Aligned Variational Autoencoders
    CVPR 2019 2019 paper | pytorch
    $\bullet \bullet$ - CADA-VAE

  69. Generative Dual Adversarial Network for Generalized Zero-shot Learning
    CVPR 2019 2019 paper | pytorch
    $\bullet \bullet$ - GDAN

  70. Hybrid-Attention based Decoupled Metric Learning for Zero-Shot Image Retrieval
    CVPR 2019 2019 paper | caffe
    $\bullet \bullet$ - DeML

  71. Generalized Zero-Shot Recognition based on Visually Semantic Embedding
    CVPR 2019 2019 paper
    Gzsl-VSE

  72. Leveraging the Invariant Side of Generative Zero-Shot Learning
    CVPR 2019 2019 paper | pytorch
    $\bullet \bullet$ - LisGAN

  73. Rethinking Knowledge Graph Propagation for Zero-Shot Learning
    CVPR 2019 2019 paper | pytorch
    $\bullet \bullet$ - DGP

  74. Domain-Aware Generalized Zero-Shot Learning
    CVPR 2019 2019 paper
    DAZL

  75. Progressive Ensemble Networks for Zero-Shot Recognition
    CVPR 2019 2019 paper
    PrEN

  76. On Zero-Shot Learning of generic objects
    CVPR 2019 2019 paper | pytorch-official
    $\bullet \bullet$

  77. Semantically Aligned Bias Reducing Zero Shot Learning
    CVPR 2019 2019 paper
    SABR-T

  78. Attentive Region Embedding Network for Zero-shot Learning
    CVPR 2019 2019 paper | pytorch
    $\bullet \bullet$ - AREN

  79. Marginalized Latent Semantic Encoder for Zero-Shot Learning
    CVPR 2019 2019 paper

  80. Compressing Unknown Classes with Product Quantizer for Efficient Zero-Shot Classification
    CVPR 2019 2019 paper
    PQZSL

  81. Gradient Matching Generative Networks for Zero-Shot Learning
    CVPR 2019 2019 paper

  82. Hierarchical Disentanglement of Discriminative Latent Features for Zero-shot Learning
    CVPR 2019 2019 paper

  83. Creativity Inspired Zero-Shot Learning
    ICCV 2019 2019 paper pytorch
    $\bullet \bullet$ - CIZSL

  84. Attribute Attention for Semantic Disambiguation in Zero-Shot Learning
    ICCV 2019 2019 paper | pytorch
    $\bullet \bullet$ - LFGAA+SA

  85. Transferable Contrastive Network for Generalized Zero-Shot Learning
    ICCV 2019 2019 paper
    TCN

  86. Rethinking Zero-Shot Learning: A Conditional Visual Classification Perspective
    ICCV 2019 2019 paper
    GXE

  87. Learning Feature-to-Feature Translator by Alternating Back-Propagation for Generative Zero-Shot Learning
    ICCV 2019 2019 paper | pytorch
    $\bullet \bullet$

  88. Modeling Inter and Intra-Class Relations in the Triplet Loss for Zero-Shot Learning
    ICCV 2019 2019 paper

  89. Dual Adversarial Semantics-Consistent Network for Generalized Zero-Shot Learning
    NIPS 2019 2019-07-12 paper
    双生 GAN；

  90. Transductive Zero-Shot Learning with Visual Structure Constraint
    NIPS 2019 2019 paper | pytorch-official
    $\bullet \bullet$ - VSC

  91. Zero-shot Learning via Simultaneous Generating and Learning
    NIPS 2019 2019 paper

  92. Semantic-Guided Multi-Attention Localization for Zero-Shot Learning
    NIPS 2019 2019 paper
    SGMA

  93. Visual Space Optimization for Zero-shot Learning
    2019-06-30 paper

    设计了一个新的 loss；

6 元学习

6.1 基于度量的

6.2 基于模型的

  1. Model-Agnostic Meta-Learning for Fast Adaptation of Deep Networks
    ICML 2017 2017-03-09 伯克利 paper | tensorflow-official | project-official | blog-official

6.3 基于优化的

6.4 其他

  1. Amortized Bayesian Meta-Learning
    ICLR 2019 paper | openreview

  2. Learning to Learn with Conditional Class Dependencies
    ICLR 2019 paper | openreview

7 应用

7.1 分类

7.2 检测

  1. Zero-Shot Object Detection: Learning to Simultaneously Recognize and Localize Novel Concepts
    ACCV 2018 2018-03-16 paper

7.3 分割

7.4 GAN

7.5 强化学习

  1. Sample-efficient policy learning in multi-agent Reinforcement Learning via meta-learning
    ICML 2019 2018-05-31 paper | openreview

TOP

附录

A 资源

  1. paper-with-code/元学习
  2. paper-with-code/少样本学习
  3. awesome-zero-shot-learning
  4. 从 CVPR 2019 一览小样本学习研究进展

B 数据集

名称 年份 数据量 类型 类别数
LAD     Large-scale Attribute Dataset 230
CUB     Caltech-UCSD Birds 200
AWA2     Animals 50
aPY     attributes Pascal and Yahoo 32
Flowers       102
SUN     Scene Attributes 717

C 比赛

  1. AI Challenger 2018
    解决方案

D 研究员

  1. Donald M. Palatucci

  2. 保罗·维奥拉
    linked in Amazon Air科学副总裁，前MIT教授，计算机视觉研究员，ICCV Helmholtz Prize 得主

  3. 奥里奥尔·温亚尔斯
    O. Vinyals google research
    Oriol Vinyals是 DeepMind 的研究科学家，主要研究深度学习，博士毕业于加州大学伯克利分校

  4. 罗布弗格斯

  5. 李飞飞
    profiles wikipedia
    斯坦福大学计算机科学系教授，斯坦福视觉实验室负责人，斯坦福大学人工智能实验室（SAIL）前负责人。专业领域是计算机视觉和认知神经科学。2016年11月李飞飞加入谷歌，担任谷歌云AI/ML首席科学家。2018年9月，返回斯坦福任教，现为谷歌云AI/ML顾问。10月20日斯坦福大学「以人为中心的AI计划」开启，李飞飞担任联合负责人。11月20日李飞飞不再担任SAIL负责人，Christopher Manning 接任该职位

  6. 杜米特鲁·艾尔罕
    profiles
    谷歌大脑研究员，目前研究重点是可扩展高质量检测、快速单次检测、图像字幕生成、视觉问答和图像理解的结构化输出

  7. 理查德·索切
    profiles Richard Socher（理查德·索赫尔）是Salesforce的首席科学家。 在此之前，他是斯坦福大学计算机科学系的兼职教授，也是2016年被 Salesforce 收购的 MetaMind 的创始人兼首席执行官/首席技术官。研究兴趣：CV、NLP、DL；

  8. 吴恩达
    profiles 斯坦福大学教授，人工智能著名学者，机器学习教育者。2011年，吴恩达在谷歌创建了谷歌大脑项目，以通过分布式集群计算机开发超大规模的人工神经网络。2014年5月16日，吴恩达加入百度，负责“百度大脑”计划，并担任百度公司首席科学家。2017年3月20日，吴恩达宣布从百度辞职。2017年12月，吴恩达宣布成立人工智能公司Landing.ai，并担任公司的首席执行官。2018年1月，吴恩达成立了投资机构AI Fund；

E 参考资料

  1. From Zero to Hero: Shaking Up the Field of Zero-shot Learning
  2. WHAT IS ZERO-SHOT LEARNING?
  3. Zero-Shot Learning
  4. 【NLP笔记】Few-shot learning 少样本学习

Comments

You May Also Enjoy