Research map / Learning theory and algorithms / Prediction and decision losses
Deep Batch Active Learning by Diverse, Uncertain Gradient Lower Bounds
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 |
|---|---|---|---|
| Use BADGE instead of the baseline active-learning algorithms | Deep neural network batch active learning · Across evaluated datasets, architectures, batch sizes, and label budgets | normalized test error | Helped |
| Replace BADGE's k-means++ sampling with k-DPP sampling (ablation) | BADGE gradient embeddings · Evaluated datasets and batch sizes | test accuracy | No clear effect |
| Replace BADGE's k-means++ sampling with k-DPP sampling (ablation) | BADGE gradient embeddings · Large batch sizes, including batch size 1000 | sampling runtime | Hurt |
| Use margin sampling instead of BADGE | Batch active learning · Batch size 10000 | normalized test error | Hurt |
| Use margin sampling instead of BADGE | Batch active learning · Batch size 100 or 1000, or MLP architecture | normalized test error | No clear effect |
| Use coreset diversity sampling instead of BADGE | Batch active learning · Large batch size (10000), ResNet, or VGG | normalized test error | No clear effect |
| Use coreset diversity sampling instead of BADGE | Batch active learning · Complex data with an MLP | test accuracy | Hurt |
| Use confidence sampling instead of BADGE | Batch active learning with gradient embeddings · OpenML #6 with MLP and SVHN with ResNet | batch diversity | Hurt |
| Use k-centers sampling instead of BADGE's k-means++ sampling | Gradient-embedding batch selection · OpenML #6 with MLP and SVHN with ResNet | batch diversity | Hurt |
| Use k-DPP sampling instead of BADGE's k-means++ sampling (ablation) | Gradient-embedding batch selection · OpenML #6 with MLP and SVHN with ResNet | batch diversity | Hurt |
Closest papers by meaning
- Batch Active Learning at Scale (NeurIPS 2021 · similarity 0.88)
- Scalable Batch-Mode Deep Bayesian Active Learning via Equivalence Class Annealing (ICLR 2023 · similarity 0.87)
- Beam Search Optimized Batch Bayesian Active Learning (AAAI 2023 · similarity 0.86)
- Towards Robust and Reproducible Active Learning Using Neural Networks (CVPR 2022 · similarity 0.86)
- Improved Algorithm for Deep Active Learning under Imbalance via Optimal Separation (ICML 2025 · similarity 0.84)
- Deep Active Learning with Noise Stability (AAAI 2024 · 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.