Research map / Label-efficient and robust learning / Recognition and long-tail
HPFF: Hierarchical Locally Supervised Learning with Patch Feature Fusion
Junhao Su, Chenghao He, Feiyu Zhu, Xiaojie Xu, Dongzhi Guan, Chenyang Si
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 |
|---|---|---|---|
| use cascade local modules instead of independent local modules only (ablation) | DGL, ResNet-32 K=16 · CIFAR-10 | test error (%): 14.08 -> 10.51 | Improved |
| combine independent and cascade levels (HiLo) (ablation) | DGL, ResNet-32 K=16 · CIFAR-10 | test error (%): 10.51 -> 9.44 | Improved |
| add Patch Feature Fusion on top of HiLo (ablation) | DGL, ResNet-32 K=16 · CIFAR-10 | test error (%): 9.44 -> 8.94 | Improved |
| add HiLo on InfoPro (ablation) | InfoPro, ResNet-101 K=4 · ImageNet | Top-1 error (%): 22.81 -> 21.35 | Improved |
| add Patch Feature Fusion on top of HiLo (ablation) | InfoPro, ResNet-101 K=4 · ImageNet | Top-1 error (%): 21.35 -> 21.14 | Improved |
| add Patch Feature Fusion to HiLo for memory (ablation) | InfoPro, ResNet-32 K=16 · CIFAR-10 | GPU memory (GB): 3.13 -> 2.31 | Improved |
| add PFF to HiLo for accuracy (ablation) | InfoPro, ResNet-32 K=16 · CIFAR-10 | test error (%): 9.59 -> 8.99 | Improved |
| apply HPFF to InfoPro on ImageNet ResNet-101 K=4 | InfoPro · ImageNet | Top-1 error (%): 22.81 -> 21.14 | Improved |
| apply HPFF to DGL VGG13 K=10 on ImageNet | DGL · ImageNet | Top-1 error (%): 35.60 -> 32.49 | Mixed |
| apply HPFF to local learning on CIFAR-10 | DGL, InfoPro, PredSim · CIFAR-10 ResNet-110 K=55 | test error (%): 14.45 -> 8.74 | Mixed |
| apply HPFF to DGL, versus BP memory | DGL ResNet-110 K=55 · CIFAR-10 | GPU memory (GB): 9.26 -> 2.44 | Improved |
Closest papers by meaning
- Scaling Supervised Local Learning with Augmented Auxiliary Networks (ICLR 2024 · similarity 0.85)
- HiGDA: Hierarchical Graph of Nodes to Learn Local-to-Global Topology for Semi-Supervised Domain Adaptation (AAAI 2025 · similarity 0.84)
- Deep Learning as Neural Low-Degree Filtering: A Spectral Theory of Hierarchical Feature Learning (NeurIPS 2026 · similarity 0.83)
- Towards Interpretable Deep Local Learning with Successive Gradient Reconciliation (ICML 2024 · similarity 0.83)
- Data-independent Module-aware Pruning for Hierarchical Vision Transformers (ICLR 2024 · similarity 0.82)
- Sparse maximal update parameterization: A holistic approach to sparse training dynamics (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.
