mlsanityData to 14 Oct 2026

Low Budget Active Learning via Wasserstein Distance: An Integer Programming Approach

arXiv 2106.02968 ↗First posted 5 Jun 2021cs.LG math.OCcited by 34

Not on the map. Why?

Abstract (arXiv)

Active learning is the process of training a model with limited labeled data by selecting a core subset of an unlabeled data pool to label. The large scale of data sets used in deep learning forces most sample selection strategies to employ efficient heuristics. This paper introduces an integer optimization problem for selecting a core set that minimizes the discrete Wasserstein distance from the unlabeled pool. We demonstrate that this problem can be tractably solved with a Generalized Benders Decomposition algorithm. Our strategy uses high-quality latent features that can be obtained by unsupervised learning on the unlabeled pool. Numerical results on several data sets show that our optimization approach is competitive with baselines and particularly outperforms them in the low budget regime where less than one percent of the data set is labeled.

Key facts. Cited by 34 papers in the map (data through 14 Oct 2026). Builds on 17 of its 52 references that are in the map. First posted on arXiv 5 Jun 2021.

Builds on

17 of 52 references.

References · 17

  1. 2015Deep Residual Learning for Image Recognitioncited by 52,096
  2. 2020A Simple Framework for Contrastive Learning of Visual Representationscited by 12,049
  3. 2013Sinkhorn Distances: Lightspeed Computation of Optimal Transportation Distancescited by 2,316
  4. 2020Improved Baselines with Momentum Contrastive Learningcited by 2,287
  5. 2020Big Self-Supervised Models are Strong Semi-Supervised Learnerscited by 1,380
  6. 2017Deep Bayesian Active Learning with Image Datacited by 819
  7. 2019Variational Adversarial Active Learningcited by 284
  8. 2013Fast Computation of Wasserstein Barycenterscited by 216
  9. 2017Deep Active Learning for Named Entity Recognitioncited by 108
  10. 2019Deep Active Learning: Unified and Principled Method for Query and Trainingcited by 93
  11. 2018Differential Properties of Sinkhorn Approximation for Learning with Wasserstein Distancecited by 56
  12. 2018Massively scalable Sinkhorn distances via the Nystr"om methodcited by 41
5 more
  1. 2019Rethinking deep active learning: Using unlabeled data at model trainingcited by 39
  2. 2017Active One-shot Learningcited by 21
  3. 2017A Meta-Learning Approach to One-Step Active Learningcited by 9
  4. 2016Learning under Distributed Weak Supervisioncited by 4
  5. 2017Scenario Reduction Revisited: Fundamental Limits and Guaranteescited by 4

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2021: 1; 2022: 7; 2023: 9; 2024: 12; 2025: 0; 2026: 5.

34 papers.

Sort:

By area:

  1. 2022Active Learning on a Budget: Opposite Strategies Suit High and Low Budgetscited by 133
  2. 2022Active Learning Through a Covering Lenscited by 66
  3. 2024A Survey on Deep Active Learning: Recent Advances and New Frontierscited by 54
  4. 2023Active Finetuning: Exploiting Annotation Budget in the Pretraining-Finetuning Paradigmcited by 36
  5. 2023A comprehensive survey on deep active learning in medical image analysiscited by 35
  6. 2022How Much More Data Do I Need? Estimating Requirements for Downstream Taskscited by 23
  7. 2023How to Select Which Active Learning Strategy is Best Suited for Your Specific Problem and Budgetcited by 17
  8. 2022Knowledge-Aware Federated Active Learning with Non-IID Datacited by 16
  9. 2024Generalized Coverage for More Robust Low-Budget Active Learningcited by 16
  10. 2021A Simple Baseline for Low-Budget Active Learningcited by 15
  11. 2022Optimizing Data Collection for Machine Learningcited by 15
  12. 2024Uncertainty Herding: One Active Learning Method for All Label Budgetscited by 15
22 more
  1. 2024Revisiting Active Learning in the Era of Vision Foundation Modelscited by 14
  2. 2023Towards Free Data Selection with General-Purpose Modelscited by 10
  3. 2024Bidirectional Uncertainty-Based Active Learning for Open Set Annotationcited by 8
  4. 2024Querying Easily Flip-flopped Samples for Deep Active Learningcited by 7
  5. 2024Active Label Correction for Semantic Segmentation with Foundation Modelscited by 7
  6. 2023You Never Get a Second Chance To Make a Good First Impression: Seeding Active Learning for 3D Semantic Segmentationcited by 6
  7. 2022Active Self-Semi-Supervised Learning for Few Labeled Samplescited by 4
  8. 2024Feature Alignment: Rethinking Efficient Active Learning via Proxy in the Context of Pre-trained Modelscited by 4
  9. 2022Pareto Optimization for Active Learning under Out-of-Distribution Data Scenarioscited by 3
  10. 2023NTKCPL: Active Learning on Top of Self-Supervised Model by Estimating True Coveragecited by 3
  11. 2023Pushing the Accuracy-Group Robustness Frontier with Introspective Self-playcited by 2
  12. 2024VeCAF: Vision-language Collaborative Active Finetuning with Training Objective Awarenesscited by 2
  13. 2024Enhancing Semi-Supervised Learning via Representative and Diverse Sample Selectioncited by 2
  14. 2024Active Prompt Learning with Vision-Language Model Priorscited by 2
  15. 2026VIOLA: Towards Video In-Context Learning with Minimal Annotationscited by 2
  16. 2026Routing, Cascades, and User Choice for LLMscited by 2
  17. 2023Pruning the Unlabeled Data to Improve Semi-Supervised Learningcited by 1
  18. 2026Scaling-Aware Data Selection for End-to-End Autonomous Driving Systemscited by 1
  19. 2023Taming Small-sample Bias in Low-budget Active Learningcited by 0
  20. 2024Dirichlet-Based Coarse-to-Fine Example Selection For Open-Set Annotationcited by 0
  21. 2026Cost-efficient Active Learning for Referring Image Segmentation and Groundingcited by 0
  22. 2026Low-Budget Active Learning through Entropic Optimal Transportcited by 0