Research map / Learning theory and algorithms / Prediction and decision losses
Making Look-Ahead Active Learning Strategies Feasible with Neural Tangent Kernels
Mohamad Amin Mohamadi, Wonho Bae, Danica J. Sutherland
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 empirical-NTK MLMOC look-ahead acquisition instead of random acquisition | Neural-network active learning · SVHN, CIFAR10, and CIFAR100 | accuracy | Improved |
| Use empirical-NTK MLMOC instead of existing look-ahead strategies | Neural-network active learning · MNIST | accuracy | Improved |
| Use sequential querying with each new true label incorporated before the next query | NTK active learning · MNIST | accuracy | Improved |
| Approximate look-ahead retraining with empirical NTK rather than naively retraining with SGD for each candidate | MLMOC · MNIST | query runtime: more than 100 times faster | Improved |
| Use empirical NTK rather than infinite NTK (ablation) | NTK look-ahead active learning · MNIST | accuracy | Improved |
| Use empirical-NTK block computation (ablation) | NTK MLMOC · MNIST | query runtime | Improved |
| Use the most likely pseudo-label rather than averaging over hypothetical labels (ablation) | MLMOC · Active learning | accuracy | No clear effect |
| Vary the maximum width of the WideResNet (ablation) | NTK active learning · MNIST | accuracy | No clear effect |
| Train with L2 loss instead of cross-entropy (ablation) | Random and NTK acquisition · MNIST | accuracy | No clear effect |
Closest papers by meaning
- Neural Active Learning with Performance Guarantees (NeurIPS 2021 · similarity 0.88)
- Neural Active Learning Beyond Bandits (ICLR 2024 · similarity 0.87)
- Improved Algorithms for Neural Active Learning (NeurIPS 2022 · similarity 0.87)
- Transductive Active Learning: Theory and Applications (NeurIPS 2024 · similarity 0.87)
- Provably Neural Active Learning Succeeds via Prioritizing Perplexing Samples (ICML 2024 · similarity 0.85)
- Active Learning Guided by Efficient Surrogate Learners (AAAI 2024 · similarity 0.85)
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.
