Research map / Label-efficient and robust learning / Test-time adaptation
Selective Classification for Deep Neural Networks
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 SR instead of MC-dropout as the confidence-rate function | Selective classification with VGG-16 · ImageNet top-1 at 60% coverage | error: 10% vs more than 20% | Helped |
| Use SR instead of MC-dropout as the confidence-rate function | Selective classification · CIFAR-10 and CIFAR-100 | risk-coverage profile | No clear effect |
| Apply SGR threshold selection to control risk using a labeled calibration sample | VGG-16 selective classifier with SR · CIFAR-10, CIFAR-100, and ImageNet | test risk relative to risk bound | Helped |
| Use RESNET50 rather than VGG-16 as the ImageNet base classifier | SGR selective classification · ImageNet top-1 and top-5 | test coverage at corresponding risk targets | Helped |
Closest papers by meaning
- SelectiveNet: A Deep Neural Network with an Integrated Reject Option (ICML 2019 · similarity 0.93)
- A Model-Agnostic Heuristics for Selective Classification (AAAI 2023 · similarity 0.86)
- Hierarchical Selective Classification (NeurIPS 2024 · similarity 0.85)
- Stratified Adversarial Robustness with Rejection (ICML 2023 · similarity 0.85)
- Confidence-aware Contrastive Learning for Selective Classification (ICML 2024 · similarity 0.84)
- Overcoming Common Flaws in the Evaluation of Selective Classification Systems (NeurIPS 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.