Research map / Label-efficient and robust learning / Recognition and long-tail
Exemplar Guided 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 |
|---|---|---|---|
| Use exemplar-guided importance sampling for initial search, then epsilon-greedy active learning | EGAL · Reddit WSD, skew 1:200, 100 labeled examples | accuracy: +11% over random search; +5% over active learning approaches | Helped |
| Use exemplar-guided search with embeddings | Active learning for Reddit WSD · Two thirds of tested words | accuracy | Helped |
| Use exemplar-guided search with embeddings | Active learning for Reddit WSD · Remaining third of tested words | accuracy | No clear effect |
| Use exemplar-guided search | Active learning on fit · Reddit WSD, fit | accuracy | No clear effect |
| Use an embedding trained for 1 or 10 epochs instead of a random convolutional embedding (ablation) | EGAL on skew MNIST · Skew MNIST | accuracy | Helped |
| Use EGAL with a poor-quality random convolutional embedding instead of least confidence (ablation) | Active learning on skew MNIST · Skew MNIST, random embedding | accuracy | No clear effect |
| Use EGAL when embeddings fail to separate classes (ablation) | Active learning on skew MNIST · Skew MNIST, random embedding | accuracy relative to standard approaches | No clear effect |
| Use exemplar-guided search at more moderate class imbalance (ablation) | Active learning for bass · Reddit WSD, bass with 50, 100, or 400 rare examples | accuracy | Mixed |
| Use importance-weighted EGAL rather than the hybrid approach (ablation) | EGAL · Reddit WSD | sampling balance | Hurt |
| Use importance-weighted search for guided search and epsilon-greedy active learning (ablation) | EGAL · Reddit WSD | accuracy | Helped |
Closest papers by meaning
- Active Learning with Partial Feedback (ICLR 2019 · similarity 0.87)
- Active Learning with Query Generation for Cost-Effective Text Classification (AAAI 2020 · similarity 0.87)
- Re-Active Learning: Active Learning with Relabeling (AAAI 2016 · similarity 0.87)
- The Power of Comparisons for Actively Learning Linear Classifiers (NeurIPS 2020 · similarity 0.87)
- A Cross-Domain Benchmark for Active Learning (NeurIPS 2024 · similarity 0.85)
- Contextual Active Model Selection (NeurIPS 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.