mlsanityData to 7 Oct 2026

Meta Discovery: Learning to Discover Novel Classes given Very Limited Data

arXiv 2102.04002 ↗First posted 8 Feb 2021cs.LGcited by 30

Not on the map. Why?

Abstract (arXiv)

In novel class discovery (NCD), we are given labeled data from seen classes and unlabeled data from unseen classes, and we train clustering models for the unseen classes. However, the implicit assumptions behind NCD are still unclear. In this paper, we demystify assumptions behind NCD and find that high-level semantic features should be shared among the seen and unseen classes. Based on this finding, NCD is theoretically solvable under certain assumptions and can be naturally linked to meta-learning that has exactly the same assumption as NCD. Thus, we can empirically solve the NCD problem by meta-learning algorithms after slight modifications. This meta-learning-based methodology significantly reduces the amount of unlabeled data needed for training and makes it more practical, as demonstrated in experiments. The use of very limited data is also justified by the application scenario of NCD: since it is unnatural to label only seen-class data, NCD is sampling instead of labeling in causality. Therefore, unseen-class data should be collected on the way of collecting seen-class data, which is why they are novel and first need to be clustered.

Key facts. Cited by 30 papers in the map (data through 7 Oct 2026). Builds on 41 of its 79 references that are in the map. First posted on arXiv 8 Feb 2021.

Builds on

41 of 79 references.

References · 41

  1. 2015Deep Residual Learning for Image Recognitioncited by 51,809
  2. 2014Adam: A Method for Stochastic Optimizationcited by 36,960
  3. 2014Very Deep Convolutional Networks for Large-Scale Image Recognitioncited by 15,267
  4. 2015Batch Normalization: Accelerating Deep Network Training by Reducing Internal Covariate Shiftcited by 6,979
  5. 2017Model-Agnostic Meta-Learning for Fast Adaptation of Deep Networkscited by 4,861
  6. 2017Prototypical Networks for Few-shot Learningcited by 3,001
  7. 2018Unsupervised Representation Learning by Predicting Image Rotationscited by 1,586
  8. 2020Meta-Learning in Neural Networks: A Surveycited by 944
  9. 2020Do We Really Need to Access the Source Data? Source Hypothesis Transfer for Unsupervised Domain Adaptationcited by 844
  10. 2015Unsupervised Deep Embedding for Clustering Analysiscited by 763
  11. 2019Generalizing from a Few Examples: A Survey on Few-Shot Learningcited by 462
  12. 2019Learning to Discover Novel Visual Categories via Deep Transfer Clusteringcited by 223
29 more
  1. 2017Positive-Unlabeled Learning with Non-Negative Risk Estimatorcited by 217
  2. 2018Learning to Propagate Labels: Transductive Propagation Network for Few-shot Learningcited by 208
  3. 2017Asymmetric Tri-training for Unsupervised Domain Adaptationcited by 152
  4. 2020Automatically Discovering and Learning New Visual Categories with Ranking Statisticscited by 141
  5. 2020Learning Deep Kernels for Non-Parametric Two-Sample Testscited by 129
  6. 2017Learning to cluster in order to transfer across domains and taskscited by 128
  7. 2016A Survey of Multi-View Representation Learningcited by 107
  8. 2019Multi-class Classification without Multi-class Labelscited by 100
  9. 2020OpenMix: Reviving Known Knowledge for Discovering Novel Visual Categories in An Open Worldcited by 93
  10. 2020Rethinking Importance Weighting for Deep Learning under Distribution Shiftcited by 92
  11. 2020Laplacian Regularized Few-Shot Learningcited by 90
  12. 2019Dynamic Instance Normalization for Arbitrary Style Transfercited by 83
  13. 2017Meta-Learning by Adjusting Priors Based on Extended PAC-Bayes Theorycited by 76
  14. 2019Hierarchically Structured Meta-learningcited by 67
  15. 2020Improving Generalization in Meta-learning via Task Augmentationcited by 55
  16. 2020What Can Be Transferred: Unsupervised Domain Adaptation for Endoscopic Lesions Segmentationcited by 50
  17. 2019Open Set Domain Adaptation: Theoretical Bound and Algorithmcited by 47
  18. 2019Few-Shot Learning with Global Class Representationscited by 44
  19. 2016Theoretical Comparisons of Positive-Unlabeled Learning against Positive-Negative Learningcited by 41
  20. 2020Part-dependent Label Noise: Towards Instance-dependent Label Noisecited by 40
  21. 2020Adaptive Task Sampling for Meta-Learningcited by 29
  22. 2020Attribute Propagation Network for Graph Zero-shot Learningcited by 29
  23. 2019Deep Model Transferability from Attribution Mapscited by 28
  24. 2019Adaptive Posterior Learning: few-shot learning with a surprise-based memory modulecited by 25
  25. 2020DEPARA: Deep Attribution Graph for Deep Knowledge Transferabilitycited by 18
  26. 2021Isometric Propagation Network for Generalized Zero-shot Learningcited by 18
  27. 2017Heterogeneous domain adaptation: An unsupervised approachcited by 17
  28. 2020Adaptive Region-Based Active Learningcited by 10
  29. 2019Butterfly: One-step Approach towards Wildly Unsupervised Domain Adaptationcited by 8

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2021: 1; 2022: 5; 2023: 16; 2024: 5; 2025: 2; 2026: 1.

30 papers.

Sort:

By area:

  1. 2022XCon: Learning with Experts for Fine-grained Category Discoverycited by 50
  2. 2023Learn to Categorize or Categorize to Learn? Self-Coding for Generalized Category Discoverycited by 39
  3. 2023CiPR: An Efficient Framework with Cross-instance Positive Relations for Generalized Category Discoverycited by 26
  4. 2023Federated Generalized Category Discoverycited by 26
  5. 2023Novel Class Discovery: an Introduction and Key Conceptscited by 24
  6. 2023Class-relation Knowledge Distillation for Novel Class Discoverycited by 23
  7. 2023When and How Does Known Class Help Discover Unknown Ones? Provable Understanding Through Spectral Analysiscited by 21
  8. 2022A Closer Look at Novel Class Discovery from the Labeled Setcited by 12
  9. 2024Robust Fast Adaptation from Adversarially Explicit Task Distribution Generationcited by 12
  10. 2024Seeing Unseen: Discover Novel Biomedical Concepts via Geometry-Constrained Probabilistic Modelingcited by 10
  11. 2022Robust Meta-learning with Sampling Noise and Label Noise via Eigen-Reptilecited by 6
  12. 2022A Method for Discovering Novel Classes in Tabular Datacited by 6
18 more
  1. 2024Exclusive Style Removal for Cross Domain Novel Class Discoverycited by 5
  2. 2025Category Discovery: An Open-World Perspectivecited by 5
  3. 2023Supervised Knowledge May Hurt Novel Class Discovery Performancecited by 4
  4. 2023Diversity-enhancing Generative Network for Few-shot Hypothesis Adaptationcited by 4
  5. 2023Generalized Category Discovery with Clustering Assignment Consistencycited by 4
  6. 2023Generalized Category Discovery in Semantic Segmentationcited by 4
  7. 2023Federated Continual Novel Class Learningcited by 4
  8. 2024Labeled Data Selection for Category Discoverycited by 4
  9. 2023Novel Categories Discovery Via Constraints on Empirical Prediction Statisticscited by 3
  10. 2023ImbaGCD: Imbalanced Generalized Category Discoverycited by 3
  11. 2023An Interactive Interface for Novel Class Discovery in Tabular Datacited by 2
  12. 2023A Practical Approach to Novel Class Discovery in Tabular Datacited by 2
  13. 2022Mutual Information-guided Knowledge Transfer for Novel Class Discoverycited by 1
  14. 2023Detecting and Learning Out-of-Distribution Data in the Open world: Algorithm and Theorycited by 1
  15. 2024When Can Proxies Improve the Sample Complexity of Preference Learning?cited by 1
  16. 2026SpaCellAgent: A Self-Evolving LLM-Based Multi-Agent Framework for Trajectory Analysiscited by 1
  17. 2021Open-Set Representation Learning through Combinatorial Embeddingcited by 0
  18. 2025ReLKD: Inter-Class Relation Learning with Knowledge Distillation for Generalized Category Discoverycited by 0