Research map / Label-efficient and robust learning / Recognition and long-tail
Towards Realistic Predictors
Pei Wang, Nuno Vasconcelos
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 |
|---|---|---|---|
| add an identically structured HP-Net with loss weighting | VGG16 classifier · ImageNet val | top-1 accuracy: 71.6% -> 73.3% | Improved |
| add an identically structured HP-Net with loss weighting | ResNet50 classifier · ImageNet val | top-1 accuracy: 76.1% -> 76.4% | Improved |
| add an identically structured HP-Net with loss weighting | VGG16 classifier · MIT67 | top-1 accuracy: 72.2% -> 72.3% | No clear effect |
| add an identically structured HP-Net | LeNet5 classifier · MNIST | accuracy: 99.0% -> 99.1% | No clear effect |
| use a ResNet50 HP-Net with a VGG16 classifier instead of a VGG16 HP-Net (ablation) | VGG16 classifier · MIT67 | top-1 accuracy: 72.3% -> 67.9% | Hurt |
| use a VGG16 HP-Net with a ResNet50 classifier instead of a ResNet50 HP-Net (ablation) | ResNet50 classifier · MIT67 | top-1 accuracy: 75.8% -> 72.7% | Hurt |
| share convolutional weights between classifier and HP-Net (ablation) | VGG16-VGG16 · MIT67 | top-1 accuracy: 72.3% -> 67.6% | Hurt |
| share convolutional weights between classifier and HP-Net (ablation) | ResNet50-ResNet50 · MIT67 | top-1 accuracy: 75.8% -> 73.2% | Hurt |
| share convolutional weights between classifier and HP-Net (ablation) | kerasNet-kerasNet · MNIST | accuracy: 99.2% -> 97.9% | Hurt |
| reject by HP-Net score instead of max softmax confidence (ablation) | F' on VGG16-VGG16 · MIT67 (reject 5%) | top-1 accuracy: 73.4 -> 75.4 | Improved |
| reject by HP-Net score instead of max softmax confidence (ablation) | F' on VGG16-VGG16 · ImageNet val (reject 5%) | top-5 accuracy: 91.2 -> 91.9 | Improved |
| reject by HP-Net score instead of max softmax confidence (ablation) | F' on VGG16-VGG16 · MIT67 (reject 25%) | top-1 accuracy: 83.1 -> 82.9 | No clear effect |
Closest papers by meaning
- When Hardness of Approximation Meets Hardness of Learning (JMLR 2022 · similarity 0.86)
- Dissecting Sample Hardness: A Fine-Grained Analysis of Hardness Characterization Methods for Data-Centric AI (ICLR 2024 · similarity 0.86)
- Overcoming Brittleness in Pareto-Optimal Learning Augmented Algorithms (NeurIPS 2024 · similarity 0.83)
- Online Classification with Predictions (NeurIPS 2024 · similarity 0.82)
- On the Hardness of Training Deep Neural Networks Discretely (AAAI 2025 · similarity 0.82)
- A Survey of Learning Criteria Going beyond the Usual Risk (Abstract Reprint) (AAAI 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.
