Research map / Label-efficient and robust learning / Recognition and long-tail
On the Learning Property of Logistic and Softmax Losses 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 in-negative-class reweighting instead of LGL | CNN2C · FMNIST | top-1 accuracy (%): 89.44 -> 91.15 | Helped |
| use in-negative-class reweighting instead of SML | CNN5C · KMNIST | top-1 accuracy (%): 96.39 -> 97.46 | Helped |
| use in-negative-class reweighting instead of LGL | CNN2C · MNIST | top-1 accuracy (%): 99.15 -> 99.29 | Helped |
| use in-negative-class reweighting instead of SML | CNN2C · FMNIST | top-1 accuracy (%): 91.15 -> 91.15 | No clear effect |
| use in-negative-class reweighting instead of SML | MobileNetV2 · CIFAR10 | top-1 accuracy (%): 91.11 -> 93.34 | Helped |
| use in-negative-class reweighting instead of LGL | ResNet-18 · CIFAR10 | top-1 accuracy (%): 91.55 -> 93.68 | Helped |
| set β to 2 instead of 1 (ablation) | LGL-INR with CNN2C · KMNIST | accuracy: 96.43 -> 96.29 | No clear effect |
| set β to 4 instead of 1 (ablation) | LGL-INR with CNN2C · KMNIST | accuracy: 96.43 -> 96.43 | No clear effect |
| adaptively upweight hard non-target classes | LGL-INR with CNN2C · KMNIST | per-class accuracy | Helped |
Closest papers by meaning
- Evaluating Robustness of Vision Transformers on Imbalanced Datasets (Student Abstract) (AAAI 2023 · similarity 0.88)
- Difficulty-aware Balancing Margin Loss for Long-tailed Recognition (AAAI 2025 · similarity 0.85)
- From Logistic Regression to the Perceptron Algorithm: Exploring Gradient Descent with Large Step Sizes (AAAI 2025 · similarity 0.85)
- Thumb on the Scale: Optimal Loss Weighting in Last Layer Retraining (NeurIPS 2025 · similarity 0.85)
- Gradient Reweighting: Towards Imbalanced Class-Incremental Learning (CVPR 2024 · similarity 0.84)
- Improved Balanced Classification with Theoretically Grounded Loss Functions (NeurIPS 2025 · similarity 0.84)
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.