Pith. sign in

REVIEW 25 cited by

Deep Batch Active Learning by Diverse, Uncertain Gradient Lower Bounds

Not yet reviewed by Pith; the record is open.

Re-run · record.json Download PDF Read on arXiv ↗

This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.

SPECIMEN: schema-true, not a live event

T0 review · schema-true

One-sentence machine reading of the paper's core claim.

pith:XXXXXXXX · record.json · timestamp

arxiv 1906.03671 v2 pith:2PUF2JX6 submitted 2019-06-09 cs.LG stat.ML

Jordan T. Ash , Chicheng Zhang , Akshay Krishnamurthy , John Langford , Alekh Agarwal This is my paper · ORCID
classification cs.LGstat.ML
keywords batchactivelearningbadgegradientalgorithmdeepdiverse
verification ladder T0 review T1 audit T2 compute T3 formal

Signed reviews

No signed human review yet.

0 comments
read the original abstract

We design a new algorithm for batch active learning with deep neural network models. Our algorithm, Batch Active learning by Diverse Gradient Embeddings (BADGE), samples groups of points that are disparate and high-magnitude when represented in a hallucinated gradient space, a strategy designed to incorporate both predictive uncertainty and sample diversity into every selected batch. Crucially, BADGE trades off between diversity and uncertainty without requiring any hand-tuned hyperparameters. We show that while other approaches sometimes succeed for particular batch sizes or architectures, BADGE consistently performs as well or better, making it a versatile option for practical active learning problems.

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 25 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. Class Balance Matters to Active Class-Incremental Learning

    cs.CV 2024-12 conditional novelty 7.0 of 10

    A distribution-matching, class-balanced selection method (CBS) improves incremental learning from unlabeled pools, beating random and standard active learning baselines on five datasets.

  2. Test-Time Coverage: Test-Conditioned Data Curation for Deployment-Aware Learning

    cs.AI 2026-07 conditional novelty 6.0 of 10

    TTCov curates training data for deployment by building an LLM-generated atomic-proposition atlas of the test distribution and greedily selecting clips that match it.

  3. ALScope: A Unified Toolkit for Deep Active Learning

    cs.LG 2025-08 conditional novelty 6.0 of 10

    ALScope provides a unified deep active learning benchmark across CV and NLP under standard, open-set, and imbalanced settings, and finds that no algorithm consistently dominates and non-standard settings remain unsolved.

  4. StepAL: Step-aware Active Learning for Cataract Surgical Videos

    cs.CV 2025-07 conditional novelty 6.0 of 10

    StepAL is an active learning method that selects whole surgical videos that are both uncertain and step-diverse, improving step recognition accuracy with fewer annotations.

  5. Exploring Active Learning for Label-Efficient Training of Semantic Neural Radiance Field

    cs.CV 2025-07 conditional novelty 6.0 of 10

    Active learning with a 3D spatial-diversity term cuts annotation cost by over 2x for semantically-aware NeRF training versus random sampling.

  6. Intended Target Identification for Anomia Patients with Gradient-based Selective Augmentation

    cs.CL 2025-06 conditional novelty 6.0 of 10

    GradSelect uses gradient signals to selectively perturb and expand circumlocution text, improving retrieval of intended target items for anomia patients.

  7. ViewPCL: a point cloud based active learning method for multi-view segmentation

    cs.CV 2025-06 conditional novelty 6.0 of 10

    ViewPCL uses Wasserstein distance between cross-view point cloud distributions as an uncertainty score, and it reports higher mIoU than ViewAL on SceneNet-RGBD.

  8. Daunce: Data Attribution through Uncertainty Estimation

    cs.LG 2025-05 conditional novelty 6.0 of 10

    DAUNCE computes training-data attribution as the covariance of per-example losses across an ensemble of perturbed fine-tuned models, reporting state-of-the-art LDS scores and the first attribution runs on proprietary LLMs.

  9. Certainty and Uncertainty Guided Active Domain Adaptation

    cs.CV 2025-05 conditional novelty 6.0 of 10

    A collaborative active domain adaptation framework that adds confident pseudo-labeled samples alongside uncertainty-based active queries, beating prior ADA methods on Office-Home and DomainNet.

  10. ALPET: Active Few-shot Learning for Citation Worthiness Detection in Low-Resource Wikipedia Languages

    cs.CL 2025-02 conditional novelty 6.0 of 10

    ALPET, an active-learning plus PET pipeline, detects citation-worthy sentences in Catalan, Basque and Albanian while needing roughly 58-72% fewer labeled examples than its CCW baseline.

  11. Deep Active Learning based Experimental Design to Uncover Synergistic Genetic Interactions for Host Targeted Therapeutics

    cs.LG 2025-02 conditional novelty 6.0 of 10

    An ensemble deep active learning framework with knowledge graph embeddings finds 92% of the top 400 HIV double-knockdown pairs after observing less than 6.3% of a 356 by 356 interaction matrix.

  12. Active Learning for Continual Learning: Keeping the Past Alive in the Present

    cs.LG 2025-01 conditional novelty 6.0 of 10

    AccuACL is a Fisher-information-based query strategy for active continual learning that improves accuracy and greatly reduces forgetting compared with standard active learning baselines.

  13. COBRA: COmBinatorial Retrieval Augmentation for Few-Shot Adaptation

    cs.LG 2024-12 conditional novelty 6.0 of 10

    A diversity-aware combinatorial mutual information retrieval objective (COBRA) outperforms nearest-neighbor retrieval for few-shot CLIP adaptation.

  14. GALOT: Generative Active Learning via Optimizable Zero-shot Text-to-image Generation

    cs.CV 2024-12 conditional novelty 6.0 of 10

    GALOT combines text-to-image generation with active learning, generating pseudo-labeled synthetic images from optimized text embeddings and reporting consistent accuracy gains over standard active learning baselines.

  15. Combining Discrepancy-Confusion Uncertainty and Calibration Diversity for Active Fine-Grained Image Classification

    cs.CV 2025-09 conditional novelty 5.0 of 10

    DECERN selects annotation samples by combining a fusion-based uncertainty score with a diversity calibration that balances closeness to uncertainty-weighted cluster centers and distance from known class anchors.

  16. Optimizing Active Learning in Vision-Language Models via Parameter-Efficient Uncertainty Calibration

    cs.CV 2025-07 conditional novelty 5.0 of 10

    C-PEAL trains the active learning selector with a loss that raises entropy for wrong predictions and lowers it for correct ones, improving sample selection for CLIP-style models under prompt learning and LoRA.

  17. Exploring Active Learning for Semiconductor Defect Segmentation

    cs.CV 2025-07 conditional novelty 5.0 of 10

    An active learning pipeline with contrastive pretraining and rareness-aware sampling reaches 98% of fully supervised mIoU on semiconductor XRM scans with only about 4.9% of labels.

  18. To Label or Not to Label: PALM -- A Predictive Model for Evaluating Sample Efficiency in Active Learning Models

    cs.LG 2025-07 conditional novelty 5.0 of 10

    PALM fits a four-parameter saturation curve to early active learning accuracy data and extrapolates it to predict the full learning trajectory.

  19. TSceneJAL: Joint Active Learning of Traffic Scenes for 3D Object Detection

    cs.CV 2024-12 conditional novelty 5.0 of 10

    A three-stage active learning sampler using category entropy, graph-based scene similarity, and mixture density uncertainty improves 3D object detection with fewer labeled scenes.

  20. Maximally Separated Active Learning

    cs.LG 2024-11 conditional novelty 5.0 of 10

    MSAL scores uncertainty by cosine similarity to fixed equiangular class prototypes, and MSAL-D adds prototype-based diversity, reporting improved AUBC on MNIST, SVHN, and TinyImageNet.

  21. Targeting Negative Flips in Active Learning using Validation Sets

    cs.LG 2024-11 conditional novelty 5.0 of 10

    RoSE, a validation-set-based filter that restricts active learning acquisition functions to estimated negative flips, improves accuracy and/or reduces negative flip rates on several image benchmarks.

  22. Improving Task Diversity in Label Efficient Supervised Finetuning of LLMs

    cs.CL 2025-07 conditional novelty 4.0 of 10

    Weighted Task Diversity allocates the annotation budget across tasks in inverse proportion to the base model's average confidence, improving MMLU and AlpacaEval scores with up to 80% fewer labels.

  23. ConBatch-BAL: Batch Bayesian Active Learning under Budget Constraints

    cs.LG 2025-07 conditional novelty 4.0 of 10

    Two budget-aware batch active learning heuristics, greedy and dynamic thresholding, reduce the number of labeling rounds needed to reach target accuracy on new building image datasets compared with random selection.

  24. ADAptation: Reconstruction-based Unsupervised Active Learning for Breast Ultrasound Diagnosis

    cs.CV 2025-07 reject novelty 4.0 of 10

    ADAptation selects breast ultrasound images for expert annotation by combining diffusion-based style transfer, hypersphere contrastive learning, and a dual uncertainty-representativeness score, and claims improved fin...

  25. Cost-effective Reduced-Order Modeling via Bayesian Active Learning

    cs.LG 2025-06 conditional novelty 4.0 of 10

    An active learning framework for reduced-order models, BayPOD-AL, shows that an error-bounded acquisition function outperforms uncertainty sampling and random sampling on a 1D heat equation surrogate task.

Pith tools