Research map / Learning theory and algorithms / Prediction and decision losses
Efficient Active Learning for Gaussian Process Classification by Error Reduction
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 |
|---|---|---|---|
| replace GPC retraining with the joint distribution integral for predictive update (ablation) | EER/SMOCU acquisition · running time, n=1000 | time per query (s): 846 -> 0.643 | Helped |
| replace GPC retraining with the joint distribution integral at small n (ablation) | EER/SMOCU acquisition · running time, n=10 | time per query (s): 0.126 -> 0.568 | Hurt |
| add importance sampling to naive Monte Carlo estimate (ablation) | acquisition function estimation · estimation accuracy, n=1000 | relative error: 0.27 -> 0.026 | Helped |
| joint distribution integral vs retraining under importance sampling (ablation) | acquisition function estimation · estimation accuracy | relative error: 0.026 -> 0.022 (n=1000) | No clear effect |
| use NR-MOCU/SMOCU with importance sampling and no retraining instead of OR-MOCU/OR-SMOCU | active learning · checkerboard 4x4 query synthesis | error regret | Helped |
| add gradient-based optimization (Adagrad) after random optimization (ablation) | NR-SMOCU · checkerboard 4x4 query synthesis | error regret | Helped |
| use SMOCU instead of MOCU as flip error rate rises (ablation) | NR-MOCU-RO · checkerboard 4x4, flip rate 0.2 | error regret | Helped |
| use NR-MOCU/NR-SMOCU instead of ADF-MOCU | active learning · checkerboard 4x4 | error regret | Helped |
| use NR-MOCU/NR-SMOCU instead of ADF-MOCU on pool-based UCI data | active learning · UCI datasets (WDBC, Ionosphere, Vehicle, Wine) | error rate | No clear effect |
| use MOCU-based acquisition instead of MES/BALD | pool-based active learning · UCI datasets | error rate | Helped |
Closest papers by meaning
- Bayesian Active Learning with Fully Bayesian Gaussian Processes (NeurIPS 2022 · similarity 0.87)
- Distributionally Robust Active Learning for Gaussian Process Regression (ICML 2025 · similarity 0.87)
- Active Learning for Gaussian Process Regression Under Self-Induced Boltzmann Weights (NeurIPS 2026 · similarity 0.86)
- GP-Tree: A Gaussian Process Classifier for Few-Shot Incremental Learning (ICML 2021 · similarity 0.85)
- Active Learning Guided by Efficient Surrogate Learners (AAAI 2024 · similarity 0.84)
- Stochastic Gradient Descent for Gaussian Processes Done Right (ICLR 2024 · similarity 0.82)
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.