Research map / Probabilistic and scientific modeling / Bayesian optimization
BatchBALD: Efficient and Diverse Batch Acquisition for Deep Bayesian Active Learning
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Jointly score acquisition points to account for dependencies within a batch | BALD for deep Bayesian active learning · MNIST, 90% accuracy | data points required: 120 / 120 / 170 -> 70 / 90 / 110 | Helped |
| Jointly score acquisition points to account for dependencies within a batch | BALD for deep Bayesian active learning · MNIST, 95% accuracy | data points required: 250 / 250 / >300 -> 190 / 200 / 230 | Helped |
| Jointly score acquisition points to account for dependencies within a batch | BALD for deep Bayesian active learning · CINIC-10, 59% accuracy | data points acquired: 1330 -> 1170 | Helped |
| Increase the number of repeated near-duplicate examples (ablation) | BALD · Repeated MNIST, 90% accuracy | data points required: 120 -> 160 -> 280 -> 300 | Hurt |
| Increase the number of repeated near-duplicate examples (ablation) | BatchBALD · Repeated MNIST, 90% accuracy | data points required: 100 data points till 90% | No clear effect |
| Use BALD instead of BatchBALD for batch acquisition | Deep Bayesian active learning · EMNIST Balanced | accuracy | Hurt |
| Use BatchBALD instead of BALD for batch acquisition | Deep Bayesian active learning · Repeated MNIST | accuracy | Helped |
| Increase BALD acquisition batch size from 1 to 40 (ablation) | BALD · MNIST | accuracy | Hurt |
| Increase BatchBALD acquisition batch size from 5 to 10 (ablation) | BatchBALD · MNIST | accuracy | No clear effect |
| Increase BatchBALD acquisition batch size to 40 (ablation) | BatchBALD · MNIST | accuracy | Hurt |
Closest papers by meaning
- Scalable Batch-Mode Deep Bayesian Active Learning via Equivalence Class Annealing (ICLR 2023 · similarity 0.91)
- Bayesian Batch Active Learning as Sparse Subset Approximation (NeurIPS 2019 · similarity 0.90)
- ALINE: Joint Amortization for Bayesian Inference and Active Data Acquisition (NeurIPS 2025 · similarity 0.85)
- Batched Energy-Entropy acquisition for Bayesian Optimization (NeurIPS 2024 · similarity 0.84)
- A Bayesian Approach to Data Point Selection (NeurIPS 2024 · similarity 0.83)
- Scalable Bayesian Learning with posteriors (ICLR 2025 · similarity 0.83)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.