Research map / Learning theory and algorithms / Prediction and decision losses
Training-Free Neural Active Learning with Initialization-Robustness Guarantees
Apivich Hemachandra, Zhongxiang Dai, Jasraj Singh, See-Kiong Ng, Bryan Kian Hsiang Low
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (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 EV-GP to select data instead of Random, K-Means++, BADGE, or MLMOC | Neural active learning · Regression tasks | test MSE | Improved |
| Use EV-GP to select data instead of the active-learning baselines | MLP neural active learning · Classification tasks | test accuracy | Improved |
| Use EV-GP data selection with convolutional architectures instead of the active-learning baselines | CNNs and WideResNets · Classification tasks | test accuracy | Improved |
| Use EV-GP to select data instead of the active-learning baselines | Neural active learning · Classification tasks | output entropy | Improved |
| Increase query batch size (ablation) | EV-GP-Emp · Classification tasks | performance | No clear effect |
| Increase query batch size (ablation) | MLMOC · Classification tasks | performance | Hurt |
| Increase query batch size (ablation) | BADGE · MNIST | performance | Hurt |
| Use a wider NN at training time (ablation) | Neural active learning · Classification tasks | model accuracy | Improved |
| Use a wider NN at training time (ablation) | Neural active learning · Classification tasks | initialization robustness | Improved |
| Increase NN width during data selection (ablation) | EV-GP · Classification tasks | model accuracy | No clear effect |
| Increase NN width during data selection (ablation) | EV-GP · Classification tasks | initialization robustness | Improved |
| Add model selection using the EV-GP+MS criterion instead of fixing a 2-layer ReLU MLP (ablation) | EV-GP · Active learning tasks where the fixed architecture is inadequate | loss | Improved |
Closest papers by meaning
- Distributionally Robust Active Learning for Gaussian Process Regression (ICML 2025 · similarity 0.87)
- Provably Neural Active Learning Succeeds via Prioritizing Perplexing Samples (ICML 2024 · similarity 0.86)
- Transductive Active Learning: Theory and Applications (NeurIPS 2024 · similarity 0.85)
- Deep Active Learning with Noise Stability (AAAI 2024 · similarity 0.85)
- Active Learning for Gaussian Process Regression Under Self-Induced Boltzmann Weights (NeurIPS 2026 · similarity 0.84)
- Active Learning for Derivative-Based Global Sensitivity Analysis with Gaussian Processes (NeurIPS 2024 · similarity 0.84)
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.
