Research map / Label-efficient and robust learning / Recognition and long-tail
Local to Global Learning: Gradually Adding Classes for Training Deep Neural Networks
Check 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 |
|---|---|---|---|
| Train with LGL rather than baseline training | VGG-16 · CIFAR-10 validation | accuracy: 0.9303 -> 0.9397 | Helped |
| Train with LGL rather than baseline training | VGG-16 · CIFAR-100 validation | accuracy: 0.7111 -> 0.7417 | Helped |
| Train with LGL rather than baseline training | VGG-16 · ImageNet | Top-1 accuracy: 71.846 -> 72.912 | Helped |
| Train with LGL rather than baseline training | ResNet-50 · ImageNet | Top-1 accuracy: 75.290 -> 76.166 | Helped |
| Use SPL instead of baseline training | VGG-16 · CIFAR-10 validation | accuracy: 0.9303 -> 0.9260 | Hurt |
| Use SPL instead of baseline training | VGG-16 · CIFAR-100 validation | accuracy: 0.7111 -> 0.703 | Hurt |
| Increase LGL minimization steps from 2 to 4 (ablation) | VGG-16 · CIFAR-10 validation | accuracy: 0.9311 -> 0.9397 | Helped |
| Increase LGL minimization steps from 2 to 4 (ablation) | VGG-16 · CIFAR-100 validation | accuracy: 0.7267 -> 0.7417 | Helped |
| Increase LGL minimization steps from 4 to 20 (ablation) | VGG-16 · CIFAR-100 validation | accuracy: 0.7417 -> 0.7491 | Helped |
| Select most dissimilar clusters instead of random clusters (ablation) | LGL · CIFAR-10 validation | accuracy: 0.9359 -> 0.9373 | No clear effect |
| Select most similar clusters instead of random clusters (ablation) | LGL · CIFAR-100 validation | accuracy: 0.7342 -> 0.7323 | No clear effect |
| Add clusters gradually through LGL rather than train on the full set from the start | DNN · Toy data | loss stability | Helped |
Closest papers by meaning
- Global Convergence and Rich Feature Learning in $L$-Layer Infinite-Width Neural Networks under $\mu$ Parametrization (ICML 2025 · similarity 0.83)
- Towards Interpretable Deep Local Learning with Successive Gradient Reconciliation (ICML 2024 · similarity 0.82)
- On the Local Complexity of Linear Regions in Deep ReLU Networks (ICML 2025 · similarity 0.82)
- Three Mechanisms of Feature Learning in a Linear Network (ICLR 2025 · similarity 0.82)
- Dendritic Localized Learning: Toward Biologically Plausible Algorithm (ICML 2025 · similarity 0.82)
- Where Do Large Learning Rates Lead Us? (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.