Research map / Label-efficient and robust learning / Semi-supervised learning
A Machine Teaching Framework for Scalable Recognition
Pei Wang, Nuno Vasconcelos
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 |
|---|---|---|---|
| add counterfactual explanations (CMaxGrad instead of MaxGrad) | MaxGrad teaching · Butterflies (MTurker labeling) | labeling accuracy: 80.33 -> 84.10 | Improved |
| add counterfactual explanations (CMaxGrad instead of MaxGrad) | MaxGrad teaching · Chinese Characters (MTurker labeling) | labeling accuracy: 81.89 -> 84.63 | Improved |
| add counterfactual explanations to MaxGrad | MEMORABLE · Butterflies (scalable recognition) | test accuracy: 74.7 -> 77.5 | Improved |
| add counterfactual explanations to MaxGrad | MEMORABLE · Gull (scalable recognition) | test accuracy: 56.3 -> 60.2 (about 4%) | Improved |
| add DivideMix noisy-label training to CMaxGrad | MEMORABLE · Butterflies (scalable recognition) | test accuracy: 77.5 -> 81.2 | Improved |
| add DivideMix noisy-label training to CMaxGrad | MEMORABLE · Gull (scalable recognition) | test accuracy: 60.2 -> 61.7 | Improved |
| add DivideMix noisy-label training to MaxGrad | MEMORABLE · Butterflies (scalable recognition) | test accuracy: 74.7 -> 78.6 | Improved |
| add DivideMix noisy-label training to MaxGrad | MEMORABLE · Gull (scalable recognition) | test accuracy: 56.3 -> 59.9 | Improved |
| use SimCLR semi-supervised model to generate counterfactual explanations (ablation) | CMaxGrad · Butterflies and Gull | test accuracy: Butterflies 77.5 -> 78.2; Gull 60.2 -> 59.7 | Mixed |
| add SimCLR-based explanations to CMaxGrad+DivideMix (ablation) | MEMORABLE · Butterflies (scalable recognition) | test accuracy: 81.2 -> 83.4 | Improved |
| add SimCLR-based explanations to CMaxGrad+DivideMix (ablation) | MEMORABLE · Gull (scalable recognition) | test accuracy: 61.7 -> 61.2 | Hurt |
| use SimCLR alone (self-supervised) instead of supervised baseline | SimCLR · Gull (scalable recognition) | test accuracy: 58.3 -> 53.0 | Hurt |
Closest papers by meaning
- Iterative Machine Teaching (ICML 2017 · similarity 0.86)
- Learning to Learn: How to Continuously Teach Humans and Machines (ICCV 2023 · similarity 0.86)
- Gradient-Based Algorithms for Machine Teaching (CVPR 2021 · similarity 0.85)
- Towards Good Practices for Efficiently Annotating Large-Scale Image Classification Datasets (CVPR 2021 · similarity 0.85)
- Data Labeling for Machine Learning Engineers: Project-Based Curriculum and Data-Centric Competitions (AAAI 2023 · similarity 0.85)
- Distantly-Supervised Named Entity Recognition with Adaptive Teacher Learning and Fine-Grained Student Ensemble (AAAI 2023 · 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.
